EXPERIMENTAL

DESIGN

Experimental Design
Cria43
1
STEERING COMMITTEE

Jaime Alberto Leal Afanador Principal

Roberto Salazar Ramos Academic Vice-Rector

HERNAN MAURICIO PULIDO

Dean of the Faculty of Basic Sciences and Engineering

Celia del Carmen Lopez

Academic Secretary Faculty of Basic Sciences and Engineering

First Edition
Copy Right

National Open and Distance University

ISBN 2005

National Center for Media for Learning

Experimental Design
Cria43
2
 FOOD ENGINEERING PROGRAM

OPEN AND DISTANCE NATIONAL UNIVERSITY

BOGOTA 2005

Elias Riaño Luna FieldEng

. Chemical. MSc.
CONTENT

Some history of experimental design

OBJECTIVES OF THE MODULE
IS EXPERIMENTAL DESIGN IMPORTANT IN APPLIED SCIENCES?
HOW WILL THE EXPERIMENTAL DESIGN COURSE BE EVALUATED?
First unit: Fundamentals of experimental design
Some interesting quotes
Not to forget
When does experimental design begin?
CHAPTER ONE: Research and experimental design
Review and evaluation questions.
1.1. Introduction to the design of experiments
1.1.1. Research purposes.
1.2. The research project.

Experimental Design
Cria43
3
1.3. Research draft.
1.4. Scientific research.
1.5. Applied research.
1.6. Research design.
1.7. The scientific method.
1.8. Research methods.
1.9. Investigative work style
1.9.1. Inductive Method
1.9.2. Deductive Method
1.9.3. Hypothetical-deductive method.
1.10. Levels of research
1.11. Stages of a research paper
1.12. Fundamentals of experimentation.
1.13. What is observation?
1.14. What is analyzing?
CHAPTER TWO: Basic definitions in the design of experiments
2.1. Design of experiments in research.
2.2. Need to design experiments.
2.3. Objectives of an experimental design.
2.4. The design of experiments in industry.
2.5. Experimental design
2.6. Utility of Statistical Design of Experiments.
2.7. Experiment
2.8. Random experiment
2.9. Experimental unit
2.11. Repeatability (precision) of an experiment
2.12. Reproducibility of an experiment
CHAPTER THREE: Review of statistical concepts
3.1. Variable.
3.2. Independent variable.
3.3. Dependent variable.
3.4. Example of independent and dependent variables
3.5. Response variable.
3.6. Factors
3.7. Level.
3.8. Variable levels
3.9. Treatment.
3.10. Witness
3.11. Replica
3.12. Randomization.
Table 1. Example of experimental units to be randomized
3.13. Activities to be developed
3.14. Answer to the questions
CHAPTER FOUR: Definitions and statistical tools applied to experimental designs.
4.1. Population and sample.
4.2. Population
4.2.1. Types of populations
4.3. Individuals or elements
4.5. Guys of sampling
4.5.1. Simple random sampling
4.5.2. Stratified sampling
4.6. Data obtained

Experimental Design
Cria43
4
4.7. Sample selection.
4.8. Parameter
4.12. Point and interval estimation.
4.13. The arithmetic mean.
4.14. Calculation of the mean from ungrouped data.
4.15. Calculating the mean of grouped data.
4.16. Variance of a population.
4.17. Standard deviation of a population
4.18. Statistical significance p.
4.8. Sample size
4.9. Example of determining the number of samples.
4.10. Recommendations for sample selection and size.
4.19. Hypothesis
4.20. Statement of a statistical hypothesis
4.21. Hypothesis formulation and testing
4.22. Test statistic.
4.23. Data analysis.
4.24. Hypothesis testing.
4.25. Z hypothesis tests for a known population mean (σ)
4.26. Example and procedure to develop for a hypothesis test itself.
4.27. t test.
4.28. Assumptions of the one-sample t test.
4.29. Z test for the difference between two means.
4.30. T test for the difference between two means
4.31. Degrees of freedom.
4.32. Analysis of variance.
4.33. Basics of variance analysis.
4.34. Least significant difference
4.35. Experimental error
4.36. Type I and Type II errors.
4.37. The most recommended methods to reduce errors
4.38. Statistical model.
4.39. Analysis of Covariance.
4.40. Recommendations
4.41. Statistical software for problem solving.
4.42. Exercise developed
4.43. Exercises and activities
UNIT TWO: Experimental designs applied to food engineering
CHAPTER FIVE: Classification and selection of experimental designs.
5.1. Characteristics of experimental designs.
5.2. Influential aspects in the selection of an experimental design.
5.3. Classification of experimental designs according to their use.
5.4. Single-factor experiments (analysis of variance)
5.5. 1. One-variable experiments
5.6. Designed unblocked experiments.
5.7. Statistical procedure for testing hypotheses in a randomized design.
5.8. Statistical model for a randomized design.
5.9. Example of an unblocked experimental design
CHAPTER SIX: Block Design
6.1. Randomized complete block design.
6.1.1. Characteristics.
6.1.2. Sources of variability.

Experimental Design
Cria43
5
6.1.3. Advantages
6.1.4. Randomization.
6.2. Statistical procedure for testing hypotheses in a randomized complete block
design.
6.3. Arranging data in a DBCA design.
6.4. Statistical model for a DBCA
6.5. Hypothesis to be tested
6.6. ANOVA for a randomized complete block design (RBD)
6.7. Comparison of treatment means in randomized complete block (RBC)
design
6.8. Example of randomized complete blocks (RBC)
6.9. Exercise to do as a group activity:
6.10. Proposed exercise
CHAPTER SEVEN: Latin Square Design
7.1. Latin square design
7.2. Formation and tabulation of experimental data in a Latin square.
7.3. Arrangement of data in a Latin Square design.
7.4. Statistical procedure for testing hypotheses in a Latin square block design.
7.5. Analysis of the Latin Square design
1.19.2 Hypothesis to be tested
7.6. ANAVA for the Latin frame design.
7.7. Example for the design of a Latin square.
7.8. Proposed exercises
CHAPTER EIGHT: Design in Greco-Latin painting
8.1. Design in Greco-Latin style.
8.2. Formation and tabulation of experimental data in a Greco-Latin square.
8.3. Statistical procedure for testing hypotheses in a Greco-Latin block design.
8.5. Hypothesis to be tested for Greco-Latin design
8.6. ANAVA for Greco-Latin design.
8.7. Example for Greco-Latin design.
CHAPTER NINE: Factorial designs
9.1. Factorial designs
9.2. Advantages of Factorial Experiments
9.3. Disadvantages of factorial designs
9.4. Factorial arrangement
9.5. Statistical model for a factorial design
9.6. Hypothesis to be tested
9.7. ANOVA for factorial (axb) or two-factor design
9.8. Example of a factorial design
9.9. Two-way factorial experiments
9.10. Factor in two-way design.
9.11. Treatment in two-way design.
9.12. Interaction in two-way design.
9.13. Steps to reach ANOVA in an experimental design.
CHAPTER TEN: 2k Factorial Designs
10.1. Coding of variables
10.2. The 2k full factorial design
10.3. Effect of a factor in a design 22
10.4. Waste analysis
10.5. ANOVA for a 22 factorial design
10.6. Example of a factorial experimental design 22
10.7. Factors and experimental domain

Experimental Design
Cria43
6
CHAPTER ELEVEN: Three-factor factorial designs
11.1. Three-factor factorial design with two levels each.
Table 38. Combination of treatments for a design 23
11.2. Effect of a factor in a design 23
11.3. Analysis of variance for a factorial design 23
11.4. Coefficients of determination
11.5. Example of a factorial design 23
11.6. Presentation of results.
11.7. Response surface methodology. Various authors.
11.8. Experimental region
11.9. Region of operability
11.10. Orthogonal polynomial.
CHAPTER TWELVE: Regression analysis
12.1. Regression analysis.
12.2. Simple linear regression.
12.3. Correlation coefficients.
12.4. Other simple linear regression models.
12.5. Multiple linear regression.
12.6. Multiple correlation.
12.7. Data transformation.
12.8. Multiple correlation.
12.9. Analysis of Covariance.
12.10. The objectives of the analysis of covariance are:
12.11. Statistical software for problem solving.
CHAPTER THIRTEEN: Other designs
13.1. Non-experimental designs. Various authors
13.2. Recommendations
13.3. Proposed exercises
13.4. Activities to be carried out for the first chapter of the second unit
APPENDIX
REFERENCES AND BIBLIOGRAPHY
Web page addresses.

Experimental Design
Cria43
7
INDEX OF TABLES AND FIGURES

Figure 1. Semantic network for experimental design in food engineering.

Figure.2. Sir Ronald Fisher.
Table 1. Contents of the first chapter of the first unit
Table 1. Example of experimental units to be randomized
Table 2. Activities to be developed for chapters one to three of the first unit
Figura 3. Example of an experimental or observational unit.
Table 4. Example of an analysis of variance
Figura 4. ANOVA decision making.
Figura 5. Cover of the Statgraphics program.
Figura 6. Excel spreadsheet *.
Table 5. Production of cape gooseberry baskets in six villages
Figura 7. aranza between the different cape gooseberry producing farms.
Table 6. Statistical summary of the total data collected on cape gooseberry
production in five municipalities of Cundinamarca.
Table 7. Statistical summary of data collected by month on cape gooseberry
production in five municipalities of Cundinamarca.
Figura 8. Chart of annual cape gooseberry production.
Figura 9. Box-and-whisker graph of monthly cape gooseberry production.
Figura 10. Graph of means and intervals for cape gooseberry production per
farm.
Table 8. Analysis of variance (ANAVA) for the example of cape gooseberry
producing farms.
UNIT TWO: Experimental designs applied to food engineering
Table 9. Statistical procedure for hypothesis testing
Table 10. Analysis of variance for treatments with a single factor, in a completely
randomized design.
Table 11. Collection and presentation of treatment results, in a completely
randomized design.
Figura 12. Flow chart for the production of doughnuts from a standardized dough.
Table 12. From randomization to the donut example
Table 13. Data obtained in the determination of hardness for the example of
donuts.
Table 14. Analysis of variance to evaluate the effect of additives in the production
of donuts.
Table 15. Homogeneity and contrast analysis of the donuts.
Table 16. Collection and presentation of treatment results, in a randomized
complete block design.
Table 17. Arrangement of the data in a randomized complete block design.
Table 18. ANOVA for a randomized complete block design
Table 19. Data obtained in the determination of solids content for the example of
tomato juice concentration.
Table 20. Processing of data obtained in the determination of solids content for the
example of tomato juice concentration.
Table 21. ANAVA for the data obtained in the determination of the solids content
for the example of tomato juice concentration.
Table 22. Compilation and presentation of treatment results, in Latin type.
Table 23. Arrangement of treatment results, in Latin table.
Table 24. ANOVA for Latin square design*
Table 25. Data collected by technologists

Experimental Design
Cria43
8
Table 26. Calculation of treatments (3rd or letters)
Table 27. ANOVA for Latin Square Design of Weight Loss Example
Table 28. Arrangement of treatment results, in a Greco-Latin table.
Table 29. ANOVA for the Greco-Latin square design*
Table 30. ANOVA for the example of a Greco-Latin square design*
Table 30. ANOVA for the example of a Greco-Latin painting*
Figure 11. A factorial experiment
Figure 12. A factorial experiment
Table 31. ANOVA for the axb factorial design
Table 32. NOVA test results for the axb factorial design
Table 33. ANOVA for axb factorial design example
Table 34. Algebraic signs for calculating the effects in a design 22
Table 35. ANOVA for a 22 factorial design
Table 36. Answers for factorial design example 22
Table 37. Matrix for factorial design example 22
Table 38. Analysis of variance for factorial design example 22
Table 38. Combination of treatments for a design 23
Table 39. Algebraic signs for calculating the effects in a design 23
Table 40. Calculating the Analysis of Variance for a Factorial Design 23
Table 41. Viscosity data of the developed beverage.
Table 42. Average and factors calculated for viscosity example
Figura 13. Steps to calculate the ANOVA in a 2k design
Table 43. Analysis of variance for developed beverage viscosity.
Table 44. Regression coefficients of the beverage viscosity data equation
developed in the previous example.
Table 45. Values for r2 , the standard error for the example of the viscosity of the
developed beverage.
Table 46. Estimated results for the viscosity data with the regression equation
found for the previous example.
Figura 16. Mean effects for viscosity example.
Figura 17. Interactions for viscosity example.
Figura 18. Interactions for viscosity example.
Table 48. Equations for estimating a simple linear regression
Table 49. Degree of association of the correlation coefficients for a set of data.
(various authors)
Table 50. Simple linear regression models
Table 51. Transformations used for data
Table 47. Activities to be developed during the academic semester.

Experimental Design
Cria43
9
Cria43 Experimental Design
11
 UNIT ONE: Fundamentals of experimental design

Experimental Design
Cria43
1
1
 Some history of experimental design

The Design of Experiments had its theoretical beginning in 1935 by Sir Ronald A. Fisher, who laid the
foundation for the theory of Experimental Design, which is now quite developed and expanded.
Currently, the applications are multiple, especially in research in natural sciences, engineering,
laboratories and almost all branches of social sciences. Experimentation provides the experimental
data, in contrast to observational data; observational data are represented as the name implies by
observations of the elementary units of a population or sample, and should not be changed or modified
by any attempt on the part of a researcher in the course of observation.

Experimental Design
Cria43
1
2
 EXPERIMENTAL DESIGN FOR FOOD ENGINEERING

SEMANTIC NETWORK FOR EXPERIMENTAL DESIGN IN FOOD ENGINEERING

Figure 1. Semantic network for experimental design in food engineering.

Experimental Design
Cria43
1
3
 OBJECTIVES OF THE MODULE

Through this module, UNAD and the Faculty of Basic Sciences and Engineering will provide the
essential elements so that students in the eighth, ninth and tenth semesters of the engineering cycle
can understand the process required to prepare a research paper supported by a statistical study,
applied to the collection and taking of experimental data produced by a properly planned methodology,
with interpretation of results until their subsequent analysis and conclusions.

Finally, the aim is for the student to be able to plan and prepare, by themselves, studies using the
application of statistics, related to those obtained through searching for documentation in libraries, in
laboratories, in centres specialising in various studies, in analysing population trends, etc., and the use
and management of some appropriate statistical software package.

Experimental Design
Cria43
1
4
 IS EXPERIMENTAL DESIGN IMPORTANT IN APPLIED
SCIENCES?

Question that has been answered by the following authors:

"Design is fundamental in the experiment. The idea expressed in the hypothesis has to be
converted into experimentation through design.

For Plutchik, "experimental designs are considered as ways of arranging the conditions of an
experiment in such a way that answers to the relevant questions are achieved."
Therefore, Kirk asserts that a design consists of the role according to which subjects are
assigned or distributed to the different experimental conditions.
When preparing the design, the experimenter establishes which variable he will act on, which are
the dependent variables that he hopes to relate to the IV, and indicates how to nullify the
negative influence of the VEs. It also establishes how many subjects the experiment will be
carried out with and decides how they will be selected." L. Gildomero Arista, Research Methodology (p. 169)

Experimental Design
Cria43
1
5
 HOW WILL THE EXPERIMENTAL DESIGN COURSE BE EVALUATED?

-Through written tests with theoretical and practical content, in which conceptual clarity, the
ability to relate concepts, adequate justification of the statistical procedures used, coherence in
the development of problem solving and practical exercises, and the correctness and adequacy
of the interpretation of the results obtained will be assessed.
- Through group work and practical exercises.

Some interesting quotes

Experimental Design
Cria43
1
6
"To experiment," says Robert Plutchik, "means 'to try out' or 'to put to the test'" (Fundamentals of
Research, 1975). What is being tested, in this case, is a hypothesis, a probable solution or
temporary response to a problem posed in science...
Let us listen to the words of Claude Bernard, a classic of the experimental method: "The name
observer is given - he says - to the one who applies simple or complex research procedures, to
make the phenomena that he makes vary and who, consequently, collects them as Nature offers
them to him.
The name experimenter is given to the person who uses simple or complex research procedures
to make natural phenomena vary or modify for any purpose, and thus make them appear in
circumstances or conditions in which Nature does not present them. In this sense, observation is
the investigation of a natural phenomenon, and experience is the investigation of a phenomenon
modified by the researcher" (Introduction to the Study of Experimental Medicine, 1960).

Experimental Design
Cria43
1
7
 Not to forget

Experimental methods are widely used in basic and applied research but with very different
purposes.

The primary goal in scientific research is usually to show the statistical significance of an effect
that a particular factor has on the dependent variable of interest.
In industry, the primary goal is to extract the maximum amount of unbiased information regarding
the factors affecting a process, production, and the quality of the finished product.
The purpose of an experimental design is to provide methods that allow obtaining the greatest
amount of valid information about an investigation, taking into account the cost factor and the
appropriate use of the available material through methods that allow reducing experimental error.

Experimental Design
Cria43
1
8
 When does experimental design begin?

The Design of Experiments had its theoretical beginning in 1935 by Sir Ronald Aymer Fisher, an
English mathematician born in London on February 17, 1890 and died in Adelaide, Australia, on
July 29, 1962.

From 1919 to 1933 he worked for the Rothamsted Experimental

Station near Harpenden, England, applying extensive statistical
analysis to agricultural research data the staff had collected.

Exercise that allowed him to develop and consolidate the basic

principles of experimental design and analysis that to date are
necessary practices to reach valid research results.
By studying and statistically analyzing experiments related to
wheat crops, he developed the analysis of variance and unified his
basic ideas on the principles of experimental design.

His first work, the article "The Arrangement of Field

Experiments" was published in 1926; in this important
article he described three fundamental components of
experiments in the area of agricultural testing: local control
of field conditions to reduce experimental error, replication
as a means to estimate the variance of the experimental
Figure..2. Sir Ronald Fisher. error and randomization to obtain a valid estimate of that
variance.

He was the one who laid the foundation for the theory of Experimental Design, which is now
quite developed and expanded.

Currently, the applications are multiple, especially in research in natural sciences, engineering,
laboratories and almost all branches of social sciences.(20)

Experimentation provides the experimental data, in contrast to observational data; observational

data are represented as the name implies by observations of the elementary units of a
population or sample, and should not be changed or modified by any attempt on the part of a
researcher in the course of observation. (30).
No other researcher has had as much impact on the statistical principles of experimental design
in his time as Ronald A. Fisher.

Experimental Design
Cria43
1
9
 Review and evaluation questions.

What is meant by research?

How is scientific research defined?
What is a research design? What is an experiment? What type of study is experimental
research?
In experimentation, what does repetitiveness mean?
Chapter 1. Research and experimental design
1.1. Introduction to the design of experiments
Experimentation plays a fundamental role in all fields of research and development. The goal of
experimentation is to obtain quality and reliable information. Information that should enable the
development of new products and processes, better understanding of a system and making
decisions on how to optimize it, as well as testing scientific hypotheses, etc.

Obviously, experimentation must be carefully planned (designed) to provide the information

sought. Such planning must consider two important aspects related to all experimentation:

1.1.2. Research purposes.

Research must fulfill two fundamental purposes:

⌧ Produce knowledge and theories, then we talk about basic or pure research.
⌧ Solving practical problems through the application of knowledge, then we talk about “applied”
or “empirical” research. This should be the field of research for UNAD schools, therefore the
policy will be to strengthen and develop applied research.

1.2. The research project.

It is a document prepared by the researcher to specify the characteristics of the investigation to
be carried out; it is generally preceded by a preliminary project.

Experimental Design
Cria43
2
0
In a project, it is necessary to complete much more information, going deeper and better defining
what is discussed in the preliminary project and adding to it what is related to the research
design, as well as a theoretical framework that makes the meaning of what is being projected
understandable. (28).

1.3. Research draft.

It is a similar but less precise document that is prepared at the beginning of the investigation, as
soon as its main characteristics have been defined.

A preliminary project must set out the characteristics of the problem, its justification, the
objectives of the research and (if applicable) the hypotheses to be verified.

1.4. Scientific research.

This type of research is based on the application of the scientific method to obtain new
knowledge that in most cases aims to move the frontiers of science.

1.5. Applied research.

This type of research is based on the application of the methodologies specific to the
development of each of the engineering fields.

1.6. Research design.

Carrying out a research design means putting into practice the general postulates of the scientific
method, planning a series of successive and organized activities where the tests to be carried
out and the techniques to be used to collect and analyze the data are found.

1.7. The scientific method.

The scientific method is the procedure or set of procedures used to obtain scientific knowledge,
the working model or general guideline that guides research.
Both basic and applied scientific research involve a process that includes several fundamental
stages, namely:

⌧ Selection, approach and analysis of the problem to be investigated. Delimitation of the topic.
⌧ Formulation of the theoretical framework (concepts – background – bibliographic references).
⌧ Presentation of the explanatory hypothesis.
⌧ Methodological design.
⌧ Data collection.
⌧ Data processing, analysis and interpretation.
⌧ Confrontation with the hypothesis.
⌧ Presentation of the final research report.
The study of the method - or of methods, if one wishes to give the concept a more general scope
- is called methodology, and includes the justification and discussion of its internal logic, the
analysis of the various specific procedures used in research and the discussion about its
characteristics, qualities and weaknesses.

1.8. Research methods.

“The word method is derived from the Greek META: along and ADOS: Path”

The method is the way of proceeding in any domain, ordering the activity to a logical end, it is a
method.

The design of a specific method, a series of successive and organized activities, which must be
adapted to the particularities of each investigation and which indicate the tests to be carried out

Experimental Design
Cria43
2
1
and the techniques to be used to collect and analyze the data.

The scientific method is based on observing some characteristic phenomena to discover the
laws that govern them.

Observation and generalization are the most important elements of the scientific method.

To give structure and foundation to the construction of the theoretical foundation of all science,
research uses methods, because they belong to the basic principles of our way of thinking and
researching, and are common to all sciences and without modifying their form or nature are
applicable to different subjects:

1.9. Investigative work style

Working styles are methods peculiar to each discipline (for example: "the anthropological
method") and the particular forms of research that are used to solve specific research problems,
as when we talk about the "qualitative method", the "experimental method" or the "statistical
method".
There is no single method of science, since the astronomer, the physicist and the economist, the
historian and the chemist, the anthropologist and the biochemist, the psychologist and the
engineer do not investigate in the same way. Historical experience also shows that scientific
procedures change because the problems that arise are different and the instruments of
assistance are in continuous evolution.

1.9.2. Inductive Method

Variant of the scientific method in which the researcher starts from the information collected
through successive observations to, through generalization, establish a law of the most universal
scope possible. (Ortiz, Uribe. Research Dictionary. Noriega editors)

1.9.3. Deductive Method

Another element of the scientific procedure is the systematic use of inference, or deductive
reasoning.
To infer means to draw consequences from a principle or assumption. Inference operates during
research and, generally, in the following manner: once a hypothesis is formulated, possible
practical consequences are deduced from it, which are then subjected, in turn, to verification.
(28).

1.9.4. Hypothetical-deductive method.

Procedure that consists of developing a theory by beginning with the formulation of its starting
points or basic hypotheses and then deducing its consequences with the help of the underlying
formal theories. (Ortiz)

1.10. Levels of research

There are different levels of research, depending on the degree of depth, rigor and accuracy and
the method used, namely:

⌧ Descriptive level: Answers the questions what is this? How does it behave? Describes the
properties or characteristics of the object under study.
⌧ Classification level: there is a systematization of the data obtained according to a previously
defined criterion.
⌧ Application level: The cause of a phenomenon or problem is sought, taking into account the
practical and theoretical context.

1.11. Stages of a research paper

Experimental Design
Cria43
2
2
The stages of any research work can be summarized as follows:

a) Problem statement.
b) General and specific objectives.
c) Hypothesis formulation.
d) Selection of experimental procedure and design.
e) Carrying out the experiment.
f) Application of statistical methods to the results.
g) Interpretation of results
h) Economic analysis and its practical usefulness for the community.

1.12. Fundamentals of experimentation.

The purpose of the experiment is:

a) The study of the variation of a population of living beings.

b) The comparison between populations and samples to judge their similarity.
e) The interpretation of results from biological, agricultural and engineering
experiments, in which populations or samples subjected to different studies or
belonging to different varieties or breeds are compared.
d) The determination of the relationship between two or more varieties
(correlation and regression).
e) The application of methods to reduce sources of error in data correlation.
f) And in segregated populations the separation of variation attributable to
selection of genes due to the environment, in studies of quantitative inheritance.

1.13. What is observation?

It is the collection of data necessary for a study. Observation is a classic method of
scientific research and is the basic way in which we obtain information about the
world around us.
It is based on the following basic principles:
⌧ It must have a specific purpose.
⌧ It must be planned carefully and systematically.
⌧ Careful control of it must be kept in writing.
⌧ Its duration and frequency must be specified.
⌧ Must follow the basic principles of reliability and validity.

1.14. What is analyzing?

After the researcher performs the proposed tests or trials, he takes data from the
dependent variable and then proceeds to categorize it, order it, and summarize it
in a way that answers the questions posed in it.

In short, it is the intellectual operation that considers the parts of a whole

separately.

Chapter 2. Basic definitions in the design of experiments

2.1. Design of experiments in research.
Designing an experiment means planning an experiment so that it gathers information relevant to
the problem under investigation.

The design of an experiment is the complete sequence of steps taken in advance to ensure that
appropriate data will be obtained so as to permit objective analysis leading to valid deductions

Experimental Design
Cria43
2
3
regarding the problem stated.

2.2. Need to design experiments.

This arises from the need to answer questions such as:
^ How will the effect be measured? Or what are the characteristics to be analyzed?
^ What factors affect the characteristics to be analyzed?
^ What are the factors that will be studied in this research?
^ How many times should the experiment be run?
^ What will be the form of analysis?
^ From what values is the effect considered important?

The design of experiments is a tool that helps us do this systematically.

2.3. Objectives of an experimental design.
Among the many objectives are:
• Provide the maximum amount of information relevant to the problem under investigation.
• The design, plan or program should be as simple as possible.
• Research should be carried out as efficiently as possible; save
time, money, personnel and experimental material. "Provide the maximum amount of
information at the minimum cost"

2.4. The design of experiments in industry.

In the case of engineering, it is essential to understand the behavior of a system, determine the
variables that affect a process, and optimize the processes and their products to reduce costs.
The design of experiments is a tool that helps to do this systematically.
In agricultural sciences, the specialist also collaborates with professionals in the field in the
planning, execution, analysis and conclusion of experiments carried out to determine the
behaviour of a breed of cattle in relation to a diet.
2.5. Experimental design
Experimental design is understood as the process of planning an experiment, such that
appropriate data is taken with the greatest possible reality, which must be analyzed using
statistical methods that derive valid and objective conclusions. We can say that the philosophy of
experimental design is to obtain information with high fidelity about the message of nature at a
minimum cost.

It is the arrangement of experimental units to control and minimize experimental error.

2.6. Utility of Statistical Design of Experiments.

The traditional method of experimentation, which perhaps arises in a more intuitive way and in
accordance with the experience of the experimenter, such as varying-a-factor-each-time, set
from initial conditions and where experiments are carried out in which all the factors remain
constant except the one being studied.

In this way, the variation in the response (Y) can be attributed to the variation in the factor, and
therefore reveals the effect of that factor. The procedure is repeated for the other factors. Method
that presents important drawbacks when there is interaction between factors.

The solution, therefore, must consist of varying more than one factor simultaneously when
carrying out a new experiment, a solution known as Statistical Design of Experiments (SDE),
universally called experimental design, as the methodology based on mathematical and
statistical aids whose objective is to help the researcher to:

Experimental Design
Cria43
2
4
1. Select the optimal experimental strategy that allows obtaining the information sought at the
minimum cost.

2. Evaluate the experimental results obtained, ensuring maximum reliability in the conclusions
obtained.

2.7. Experiment
Set of rules used to obtain a sample of the population and at the end of the test obtain
information about the population.
An experiment is a procedure used to test (confirm or verify) one or more hypotheses related to a
given phenomenon, by observing and measuring the variables that influence it.
For example, all the laboratory tests and field tests that you carry out to develop your degree
work.

It is a change in the operating conditions of a system or process to obtain a sample of the

population and, at the conclusion of the test, obtain information about the population or the
product obtained.

For example, varying the operating conditions (temperature, pressure, agitation speed of a
process, rations for cattle, doses of agrochemicals).

Or the use of different proportions of raw materials and additives to improve the condition of a
mass consumption product.

Experimentation is one of the steps of the scientific method.

Examples of experimental systems are:

- A chemical and/or biochemical reaction, whose yield (Y) may depend on, among others, the
reaction time (t1), the reaction temperature (T2) and the type of microorganism (Mo1) used. Other
variables that may influence are, for example, the presentation of the substrates, the agitation
speed,.............................................................................................................

2.8. Random experiment

Activity that results in or produces an event. Test where there are two or more possible
outcomes, and it cannot be anticipated which of them will occur.
When in your tests you cannot completely control all the variables involved in the process such
as internal variations in a food diet, raw materials, ambient temperature, metabolism of the
animals involved in the experiment...

2.9. Experimental unit 1 10

It is the experimental material to which a treatment is applied uniformly.

It is the variable that the experimenter manipulates or unit of analysis, characterized by attributes
that differentiate it from one another partially or totally; they can be subjected to ordering
according to some criterion.
Experimental unit is the material (object of the experiment) to which a treatment is applied
uniformly.

It is also the unit of observation, it is the entity (e.g. food, plot, livestock, taster) on which one or
more characteristics of interest are measured.

For example: It can be a product to be improved, a set of raw materials (fruits, vegetables,

Experimental Design
Cria43
2
5
muscle tissues), a batch of a product, a test tube, etc.

A bakery product, a variety of agricultural products (corn, wheat flour), fruits from a region,
muscle tissues, a substrate for biotechnological trials.

2.10. Selection of experimental material. (8)

Sampling, or selection of the experimental units that will make up the sample to be analyzed,
must be designed so that this sample is representative of the phenomenon or population studied,
so that the results obtained can be extrapolated to situations different from those to be
measured.

The main problems that compromise this representativeness are the lack of independence of the
samples and the effect of uncontrolled factors that may be affecting the results.
The choice of units is important; for example, in planning field experiments, numerous studies
have been made of the variability between yields of raw materials from crops grown in plots of
different sizes and shapes and from companies or brands, under uniform treatment.
If the results of the experiment are to be applied to unselected material, these types of
specialization have potential disadvantages.
The responses obtained for treatments on highly selected experimental material may not be the
same as those obtained from non-selected material.
It is important to define the uniformity criterion which refers to the treatment, the experimental
material and the technique.
The amount of sample should be sufficient to allow the experimenter to make auxiliary
measurements.

2.11. Repeatability (precision) of an experiment

It is the variation in measurements made by a single operator on the same unit and with the
same measuring instrument or equipment.

It is defined as the variation around the mean.

This variation should be small with respect to the specifications and/or the process variation.

2.12. Reproducibility of an experiment

Variation between the means of measurements made by several analysts with the same
experimental units and the same measuring instrument.

Chapter 3. Review of statistical concepts

3.1. Variable.
They can be defined as everything that we are going to measure, control and study in an
investigation or study.
They are properties that are manipulated, measured and controlled during the course of an
investigation. They differ from each other in many aspects, the most important being the role
they play in an investigation and the type of measurement applied.
A variable can also be defined as anything that can take on different values, from a quantitative
or qualitative point of view.
Therefore, it is important, before starting an investigation, that we know what variables we are
going to measure and how we will do so. That is, all variables involved in the experiment must be
susceptible to measurement.

Experimental Design
Cria43
2
6
For example, time is always considered a variable, the diameter that defines the size of an
agricultural product, a process temperature, a substance to be added to a formulation or ration,
the type of support to be used to immobilize an enzyme, a fertilizer; etc.
Cubillos, Munca, Jose Miguel. Statistics 1 Self-Training Module. Higher School of Public Administration Territorial Public Administration
Program.2002.
3.2. Independent variable.
An independent variable is one that, within the established relationship, does not depend on any
other, although it could be dependent if we were studying another problem.

They are the conditions manipulated by the researcher in order to produce certain effects.

It is the truth value given to a hypothesis in relation to the cause,

3.3. Dependent variable. (29 and others)

The independent variable is that property of a phenomenon that will be evaluated for its capacity
to influence, affect or affect other variables.
It is the factor that the researcher observes or measures to determine the effect of the
independent variable.
The dependent variable is the response variable or output variable.

3.4. Example of independent and dependent variables

For example, if the research engineer is going to test the hypothesis that by supplying a certain
amount (%) of yeast to a dough, the degree of hardness of the final product increases. In this
case, the independent variable will be represented by the amount of yeast manipulated by the
experimenter, and the dependent variable will be the degree of hardness of the products
obtained.

3.5. Response variable.

It is a variable that the experimenter measures after carrying out the treatments to see how it is
affected by the experimental variable.
For example, when feeding a ration to a litter of rabbits and quantifying the weight gain of each of
them after two months.

3.6. Factors
They are the process variables in an investigation, which can be measured on a continuous
scale. For example pressure, temperature, weight, concentration.

3.7. Level.
A specific value within a scale, either qualitative or quantitative, that takes the experimental
variable selected for the study.

3.8. Variable levels

They are the different values (manipulated according to what is desired) that are assigned to
each factor studied (independent variable) in an experimental design.

The variation (manipulation) of an independent variable can be carried out in two or more
degrees. Levels can be taken from a universal measurement scale such as two degrees Celsius,
10 grams or %; etc.

The group with no manipulation is called the control group and the group with the presence of
the independent variable is the experimental group.

In other experiments the independent variable can be manipulated to more than two degrees; for

Experimental Design
Cria43
2
7
example when three levels are used and represented as follows:

X1= permissible concentration of a preservative (for purees, meat emulsions, sauces, rations,
agrochemicals; etc.)

X2 = intermediate concentration of the same preservative.

X2 = maximum concentration of the same preservative.

X0 = absence of preservative or purees, meat emulsions, sauces, etc., in their natural state.

A combination of all the factors is called treatment.

3.9. Treatment.
Treatments constitute the different procedures, processes, factors or materials whose effects are
going to be measured and compared.

The treatment establishes a set of experimental conditions to be imposed on an experimental

unit within the confines of the selected design.

In a univariable experiment, each treatment is a level of the experimental variable.

Example:

Dose of acidifier, protein, proportion of stabilizer, origin of agricultural products, varieties of a

given crop.

The dose supplied (%) of a substance to improve the conditions of a food; the applied substrate
(different % of nutrients) in a biotechnological process; the cooling speed (refrigeration); the
varieties of an agricultural product (tangelo orange, valencia, mangoss, etc.); the composition of
a storage atmosphere (% of gases), etc.

In an experiment a combination of several simple factors can be considered, for example

composition, temperature, pressure, time, etc.

3.10. Witness
The control is the additional comparison treatment, which should not be missing in an
experiment; for example, if five treatments with acidulant are used, the control can be the
treatment that does not include acidulant. The choice of the control treatment is of great
importance in any research, as it constitutes a reference for the experiment and serves to
compare the treatments being tested.
This is the additional comparison treatment, which should not be missing in an experiment.
For example, if you want to test the degree of gelation in a jam by adding a new pectin, or the
extraction yield of an orange juice by using a new extraction technique, the test will be planned in
such a way that a jam with a well-known traditional pectin is included and the extraction will
include the traditional technique that was being used.

3.11. Replica
Each repetition of a treatment is called a replicate.

A complete repetition of a treatment is a replicate.

The replica produces the following effects:

Experimental Design
Cria43
2
8
1. Typically, the average of several replicates of a treatment comes closer than any single test
to the "true" effect of that treatment.

2. By observing the variation between replicates of the same treatment, it is possible to estimate
the magnitude of the experimental error. This makes it possible to determine whether the
differences observed between treatments in a response variable are really caused by the
treatments or are only due to the interfering variables.

3. By randomly assigning experimental material to different replicates, it is possible to reduce

biases that could produce misleading results.

3.12. Randomization. (8)

It is a technique within the experimental design which has as its primary purpose to assign the
experimental material an order to carry out the individual runs or tests of the experiment. This
order is carried out at random. It helps to average out the effects of extraneous factors that may
be present.

If the number of units does not exceed 16, tables from statistics books can be used.

For example, suppose there are 3 treatments, 2 of which have 4 replicates and the third 8.

In any convenient order, the numbers 1 through 16 are assigned to the units.

And according to TABLE 1. From chapter one of unit one (Random permutations), permutations
of: 5 for a random permutation we select; for example: the column ( ) which has the
following order:

4 3 9 2 9 (for the first five highlighted columns and for example for U12 - U16 96981

And we began to give these codes to the units starting with U1 and so on until ending with U16.

Table 1. Example of experimental units to be randomized

U1 U2 U3 U4 U5 U6 U7 U8

U9 U10 U11 U12 U13 U14 U15 U16

*** For example, an experiment could be done to find out how different types of sugars affect texture or how processing temperatures
affect the moisture content of a product. The information obtained makes it possible to manipulate one or more variables in order to control
others.

3.14. Answer to the questions

What is research? According to Enrique Biermann, the research is based on different concepts, among which the following can be
summarized:
"It is the process by which man, starting from questions of diverse order and importance, seeks to obtain answers, to achieve different
objectives"
“It is a process of theoretical and practical knowledge and learning, which is developed in different phases, through the use of different
instruments or tools”
What is research? According to Enrique Biermann, research obeys different concepts, among which the following are summarized: "It is the
process by which man, starting from questions of diverse order and importance, seeks to obtain answers, to obtain different objectives"
“It is a process of theoretical and practical knowledge and learning, which is developed in different phases, through the use of different
instruments or tools”

Experimental Design
Cria43
2
9
In summary, we can say that: "to investigate is to look for answers to different unknowns that arise in the face of any event or problem" but, as
we find answers, other unknowns arise, which forces us to continue investigating.
So what is scientific research?
Scientific research can be defined as a type of "systematic, controlled, empirical and critical research of hypothetical propositions on
presumed relationships between natural phenomena." Systematic and controlled implies that there is a method and that the facts are not left
to chance; “empirical” implies that it is based on observable phenomena of reality; and “critical” implies that the facts are presented objectively,
leaving aside personal preferences or value judgments.

Experimental Design
Cria43
3
0
TABLE 2. Activities to be developed in chapter three

Activities Delivery date Learning phase

Activities Work to be done

Each student must in a According to the

series of proposals schedule agreed
Recognition
Activities No. 1 investigation with the tutor
Work to be done identify who the population
by students is, the sample and identify
what type of variables are
included.

Activities No. 2 For a degree project According to the

developed in the schedule agreed
professional cycle, carry out with the tutor
statistical research on the Deepening
characteristics of the raw
materials used.

Practical exercise And according to the According to the

of study, permutation table as schedule agreed
reflection and would perform the with the tutor
analysis. stabilization of a product
Number 2.37 developed in the PEDT? ***
Or from a practice that you
have done in a course taken
in the profession you study.

According to the
Activities No. 3 schedule agreed
Exercise proposed by the with the tutor
course teacher
Work to be done
by students Transfer
Chapter Four: Definitions and tools applied to experimental designs
4.1. Population and sample.
Population or universe is any set of units or elements such as people, farms, agricultural product,
finished product, etc., clearly defined for which estimates are calculated or information is sought.

In every experimental design, the units, their content and extension must be defined.

When it is impossible to obtain data from the entire universe, it is advisable to extract a sample, a
subset of the universe, that is representative.

In every project the size and type of sampling to be used must be specified:

4.2. Population
There are several definitions for a population, we will transcribe some of them:

It is the set formed by all the units that are the object of a statistical study.

We can also define it as the set of individuals or elements that fulfill certain properties
common.

Or as the collection of all the elements that are being studied and about which we are trying to
reach conclusions.

Experimental Design
Cria43
3
1
4.2.1. Types of populations
In relation to population size, this can be:

• Finite, as is the case with the number of people who make up a school, a tasting panel,
the trees on a farm, the cattle on an experimental farm, etc. Also all products obtained in
one day of processing (three shifts).
• Infinite, if for example we study the random mechanism that describes the sequence of
heads and tails obtained by repeatedly tossing a coin, the cattle that can be obtained
worldwide, the production of a metabolite by certain microorganisms in the Amazon
rainforest.

Let us consider that a sensory tasting of a new food is going to be carried out in the population
formed by all the students of the school (foods+systems+basic+pharmacy regency) or finite
population.
4.3. Individuals or elements

People or objects that contain certain information that is desired to be studied.

For example, all sausages from a work shift in which some stages of production have been
modified.

4.4. Sample
Representative subset of a population.

It may be the sausage(s) taken from a production for sensory and physicochemical analysis.

Observational
unit

A cattle from a herd, a quail from a poultry farm.

Figure 3. Example of an experimental or observational unit.

4.5. Types of sampling

4.5.1. Simple random sampling

In this type of sampling each element has the same probability of being chosen and all
combinations are equally probable. The population is assumed probable.
The selection can be made with or without replacement; in the first case each element can be
selected more than once while in the second it can be selected only once.

With replacement, when a packaged sample (one pound) is taken in a storage study of an
edible, tests are performed and the sample is returned to the storage site.

Experimental Design
Cria43
3
2
4.5.2. Stratified sampling
In this case the population can be divided into groups or categories (strata). For example, there
is a population of 200 cattle for a parasitism study, and for this study they are classified
according to breed (20 Norman, 30 Santa Getrudis, 100 Holstein, 50 Jersey), age, and any other
criteria that the researchers consider important.

After dividing the samples, it becomes a simple random sampling.

4.6. Data obtained

These are the reports obtained as a result of the research activity and provide knowledge of the
variability of the experiment's observations.

These are usually the ones reported as Yi.. ; etc. In all the exercises that will be carried out in the
module.

4.7. Sample selection. (Various authors)

For a sample to be representative of the entire population, it must be large enough to reflect the
variation within the population.

For example, to evaluate the juice yield per hectare of a tangelo orange production planted on an
area of 30 hectares, one could measure the yield of one hectare and multiply it by the number of
hectares of the site from which they come.

However, the yield between trees varies considerably; a more accurate estimate would
undoubtedly be obtained by measuring the juice yield of, for example, 10 hectares, and obtaining
a result that is more credible and closer to that obtained by measuring all the plants.

4.8. Parameter
This is a term used by statisticians to describe the characteristics of populations and samples. It
refers to measurements of central tendency and variation, such as the mean and standard
deviation, of the fixed and invariable characteristics of populations. That is, it is a function defined
on the numerical values of measurable characteristics of a population.

4.9. Statistical
It is a function defined on the numerical values (observations) of a sample.
It can be the mean, the variance, etc. Table 3.

4.12. Point and interval estimation. 30

A point estimator of an unknown parameter is a statistic that generates a single numerical value
that is used to estimate the value of the unknown parameter. For example, three parameters
related to the quality characteristics of a process, about which one frequently wishes to make
inferences, are: The process (population) mean, the variance a2 or standard deviation a of the
process.

4.13. The arithmetic mean.

It is the average result of a set of values, some of which have a more representative frequency.

When we refer to the "average" of something, we are talking about the arithmetic mean.

To find the arithmetic mean, we add the values and divide the result by the number of
observations.

Experimental Design
Cria43
3
3
The mean of a population is symbolized by µ (Greek letter miu).

The number of elements in a population is denoted by the italic capital letter N.

Statistically, Latin letters are generally used to symbolize information about samples and Greek
letters are used to refer to information about populations.

Table 3. Symbols for parameters and statistics

FEATURE PARAMETER SYMBOL STATISTICAL SYMBOL

Average
x
Standard deviation μ
σ SD

Variance 2σ 2 S2

Correlation
ρ r

Proportion
π p

4.14. Calculation of the mean from ungrouped data. (2, 3, 7)

Population average:

µ = x / N Equation 1
X = x / n Equation 2

To calculate this mean, we add up all the observations.

Statisticians refer to this type of data as ungrouped data.

4.15. Calculating the mean of grouped data.

To calculate this statistic, the following considerations must be taken into account:

⌧ A frequency distribution consists of data grouped into classes.

⌧ Each value of an observation falls into one of the classes.

⌧ We do not know the individual value of each observation.

⌧ From the information in the table, we can easily calculate an estimate of the value of the
mean of this grouped data.

Experimental Design
Cria43
3
4
⌧ To find the arithmetic mean of grouped data, we first calculate the midpoint of each class.

⌧ To ensure that the midpoints remain in rounded figures, we round the amounts.

⌧ We then multiply each midpoint by the frequency of observations in that class, add up all the
results, and divide this sum by the total number of observations in the sample.

X = ∑ (fX /n) Equation 3

f = frequency of observations of each class

x= midpoint of each class of the sample
n = number of observations in the sample
4.16. Variance of a population. (2, 3, 7 and various authors)
It is equal to the sum of the squares of the differences around the population mean μ , divided
by the population size.

N
σ2
=∑(Xi-μ) 2
N
i=1
Equation 4

∑
N

(Xi - μ) 2
= Sum of the squares of the differences between the values Xi and μ i=1

Equation 5
N = population size

4.17. Standard deviation of a population

The population standard deviation is:
N

σ=
V[∑(Xi - μ) 2
[N] Equation 6
i=1

4.18. Statistical significance p.

The observed significance po value of a statistical test is the smallest value of α] at which H0

can be rejected. It is the real risk of committing a type I error, if H0 is rejected based on the observed
value of the test statistic. In other words, the p-value measures the strength of the evidence
against H0.

In experimentation, it is customary to use significance levels of 5 and 1 percent.

If the p-value is less than a preassigned significance level α], then the null hypothesis can be
rejected, and the results reported as statistically significant at the α level.

When the null hypothesis is true, then the difference is said to be significant. When reporting
statistical significance, many researchers write ( p < 0.05), to indicate that the p value of the test

Experimental Design
Cria43
3
5
was less than 0.05, making the results significant at the 5% level. In journals that publish
research papers this is expressed as significant (p<0.05) and the value of the test criterion is
marked with an asterisk *.

If a value of the test criterion has a probability less than 1 percent when the null hypothesis is
true, the difference is said to be highly significant and the calculated value of the test criterion is
denoted by two asterisks **.

4.8. Sample size

In an experiment, taking a smaller sample than necessary will naturally be less costly but will
lead to results without practical use since the standard error increases to unacceptable levels.

A wide variety of formulas are used to calculate the required sample size, while a variety of
computer programs perform sample size calculations for a wide range of study designs and
statistical methods. Lerman (1996) provided an advanced description of the logic behind
calculating sample size for engineering research.

In any experimental design, the number of replicates to be performed for each treatment must be
decided. This number of replicates determines the sample size.

For example, if small differences are expected to be found in the final response, the sample
number should be as large as possible to detect significant differences. And vice versa if the
differences are large.

And on the contrary, if there are several treatments, the replicas should be reduced.

It is also very important to consider the cost, the material available and the availability of the
equipment to be used.

Hence, it is very important to have criteria to calculate, even approximately, the sample size. For
this purpose, formulas based

Experimental Design
Cria43
3
6
 on the properties of the normal distribution and the standard error are used, which allow the
calculation of confidence intervals to obtain a certain probability P. To determine the sample size,
reliability, probability and sampling error must be considered.

For example, be it the average a population and x, the sample mean

interval width is defined as; of
Interval width = μ- X Equation 7

that is, the deviation of X (sample mean/) with respect to μ (population mean).

μ- X = (ZS) / 7n Equation 8

Yes YesLμ-X =dS, we can say that

d=ZS Equation 9

=ZS /d
Equation 10
n = Z2S2 I d2

Among the many formulas proposed by statisticians to calculate the size of a sample, the following is
obtained if the researcher plans to carry out sampling with replacement or from an infinite population,
the sample number to consider is:

n= Z2 S2N
/ (N-1)+
d2 Z2 S2
Equation 11

N = Population size

S = Range of the variable (RV) / 6 = {[Maximum value of the Variable - minimum value of the

Variable] / 6} Equation 12
Where Z = Reliability coefficient and depends on the desired probability or critical value corresponding
to an area of (1-α )/2 from the center of a normal distribution.
S= Standard deviation of the sample.
n = Sample size.

The above formula requires that the standard deviation be known, which is not known and must be
estimated, for which the following procedures can be used:

- A pilot sample is taken from the population. The variance of this sample is used as an estimate of S 12
3
.
- If data from previous similar studies exist, one of them can be used as an estimate.

1 What degree of significance ( α

degree or p value) do you require in relation to the null hypothesis?
2 What is the degree of the power (equal to 1 or value of r2 ) that you want?
3 How large must the difference between the mean and the standard value or norm be for the difference
to be significant from a food engineering perspective?

Experimental Design
Cria43
3
7
4.9. Example of determining the number of samples.
a) Infinite population
In a canning factory, a sample of 700 jams is established and a probability that some of them present
hysteresis (weeping) is 30%, with a reliability of 95%, find the sampling error.

Since the variance is not known, suppose that a series of samples (20 bottles) are taken at random in
two process shifts and it is determined that they have a standard deviation of S2 = 10
The next step is to calculate the reliability coefficient Z associated with the 95% reliability.
Z = (1-α)/2 = (1-5)/2= 2

Applying the formula n= Z2 S2N

/ (N-1)+
d2 Z2 S2

What order would be the sample size to be taken?

And the sampling error if only 10 bottles had been taken?

b) Calculate the sample size for the same product (marmalade) with a reliability of 96%, an error of
1.5%, and a probability of the phenomenon occurring estimated at 60%.

n = Z02 PQ / E2

n = 9 (60)(40)/ 2.25 = 864 bottles

4.9. Recommendations for sample selection and size. (8 and various authors)
Researchers must know how large a sample is needed before beginning an investigation, otherwise
they will not be able to determine significance when it arises or is required.

To calculate the sample size to use in a research that requires a mean, the following four questions
must be answered:

4. What is a sufficient estimate of the standard deviation in the population?

4.19. Hypothesis
Etymologically, it is the assumption of a truth that must be verified or rejected.

It is an explanation given to a fact at the beginning of an investigation; it is a conjecture about reality.

And it serves to guide the researcher in finding a truth. ( See the work module for the technological cycle and
professional cycle of the same compiler)

4.20. Statement of a statistical hypothesis

A statistical hypothesis is a statement about the values of the parameters of a population or process,
which is capable of being tested based on the information contained in a representative sample
obtained from a population. For example, the statement "this process produces less than 6% defectives"
can be stated statistically, in terms of, the unknown proportion p of defective items that the process
generates.

Ho : p =0.06 (the proportion of defectives is 0.06)

HA: p < 0.06 (the proportion is less than 0.06)

4.21. Hypothesis formulation and testing
A hypothesis is a statement about something.

Experimental Design
Cria43
3
8
In statistics, it can be a guess about the value of an unknown parameter.

The test is the means of verification to know if something is true or false and to what degree we can say
that it is true or false.

Steps in hypothesis testing:

⌧ Define the null hypothesis: assume a hypothesis about a population.

⌧ Formulate an alternative hypothesis: it is a counter-hypothesis.
⌧ Define a decision criterion to reject or not the null hypothesis.
⌧ Inspect the sample data.
⌧ Calculate a sample statistic.
⌧ Use sample statistics to evaluate the hypothesis.

Recommendation should be to "do not reject" a hypothesis rather than "accept" it, since the evidence is
not conclusive.

4.22. Test statistic.

Once the hypothesis is raised, a random sample is taken (or data is obtained through an experiment
planned according to the hypothesis of the population under study). The test statistic is a number
calculated from the data and the null hypothesis, the magnitude of which allows us to determine whether
the null hypothesis Ho is rejected or accepted. The set of possible values of the test statistic that lead to
rejecting Ho is called the rejection region or interval for the test, and the possible values where Hl is not
rejected are called the acceptance region or interval.

For example, for the hypotheses posed the test statistic is given by 0.08 Zo = vlO.08(1- 0.08) / n'

4.23. Data analysis. (2, 3, 7 and various authors)

The analysis depends on the level of measurement of the variables, the way in which the hypotheses
have been formulated, and the interest of the researcher in the problem being investigated.

We can use a series of numbers known as summary statistics to describe the characteristics of the data
set. Two of these characteristics are of particular importance to decision makers: central tendency and
dispersion, among which we have the median, mode, measures of central tendency, mean deviation,
dispersion, skewness in curves and graphs, kurtosis, arithmetic mean, median, mode, dispersion,
frequency distribution, histograms, population variance, Z scores, ratios and cups, analysis of variance.

4.24. Hypothesis testing. (2, 3, 7 and various authors)

In general, a hypothesis test begins with a theory or assertion regarding a certain parameter of a
population for which two hypotheses known as:

• The Null Hypothesis Ho which is the hypothesis that is always tested.

• And the alternative hypothesis H1 is stated as the opposite of the null hypothesis and represents
the conclusion that is supported if the null hypothesis is rejected.

In what is known as classical hypothesis testing methodology, the following points are recommended to
be taken into account:

• The null hypothesis Ho always refers to a specific value of the population parameter (such as
not to the master statistic (such as X-).
μ),
From this section on, you should consult the tables found in statistics texts such as the T-Student tables,

Experimental Design
Cria43
3
9
F distribution tables, Duncan comparison of means, etc.

4.25. Z hypothesis tests for a known population mean ( σ)

Equation 13
X = Observed sample mean; μ = hypothesized mean
The denominator represents the standard error.
= Population variance
σ2

4.26. Example and procedure to develop for a hypothesis test itself.

A consumer organization is interested in determining whether there is a difference in weight between
different brands of 500-gram boxes of breakfast cereal, for which it accepts a variance of 10 grams.

For this purpose, the student (if possible) will collect 25 pieces of data, store them in a table and analyze
them; it is advisable to work with a significance level of 5%.

The steps to follow for hypothesis testing are as follows:

1. State the null hypothesis Ho. This must be expressed in statistical terms.

For example: When testing whether the average filling amount is 500 grams, the null hypothesis states
μ
that ( ) is equal to 500 grams.

2. State the alternative hypothesis H1. It must also be expressed in statistical terms.

When testing whether the average filling amount is 500 grams, the alternative hypothesis ensures that (
μ ) is less than 500 grams.
3. Choose the significance level α
. This is determined after taking into account the specified risks of
committing type I and type II errors in a particular situation.

The company chose α= 0.05.

4. Choose sample size n. This is determined after taking into account the specified risks of committing
type I and type II errors (i.e., the selected levels of α and β) and considering the budget constraints
when conducting the study.

In this case, 25 randomly selected cereal boxes were weighed.

5. Determine the appropriate statistical technique and corresponding test statistic to use. Since σ is
known because the company specified it was 15 grams, a Z test was chosen.

6. Set the critical values that divide the rejection and non-rejection regions. Once the null and
alternative hypotheses have been specified and the significance level and sample size determined, the
critical values for the appropriate statistical distribution can be found so that the rejection and non-

Experimental Design
Cria43
4
0
rejection regions can be indicated.

In this case, the values + 1.96 and - 1.96 will be used to define the regions because the Z statistic refers
to the standard normal distribution.

7. Collect the data and calculate the sample value of the appropriate test statistic.

Take a data collection (determination) and calculate the mean X = ........................ grams, then and get
the value Z = + .........................................................................................................

8. Determine whether the test statistic is in the rejection or non-rejection region. The calculated value of
the test statistic is compared to the critical values of the appropriate sampling distribution to determine
which region it lies in.

In this case, . Z= +...........is in the rejection or non-rejection region because - 1.96 < Z == + < +
1 .96.

9. Make a statistical decision. If the test statistic is in the non-rejection region, the null hypothesis, Ho,
cannot be rejected; if the test statistic is in the rejection region, the null hypothesis is rejected.

10. Express the statistical decision in terms of a particular situation.

If the difference between the quantities is affirmative, what corrective action would you propose and
apply?

4.27. t test. (12)

The t test is sometimes known as the Student t test, named after the person who first studied it in 1890 .
Student was actually a mathematician named William Gosset, employed by the Guinness Brewery, who
was forced to use the pseudonym Student because company policy prohibited employees from
publishing their research.

Gosset discovered that when an observation comes from a normal distribution, the means are
distributed normally, only if the true standard deviation of the population is known.

The t test is widely used in all areas of science.

The t distribution is similar to the z distribution which was explained in the previous section and one of
its main uses is to answer research questions about means and when the true standard deviation is
unknown.

t = X - μ(S n) Equation 14

S, is the standard deviation calculated by:

n
S=V[∑(Xi-X) n-1]
2

i=1
Equation 15

4.28. Assumptions of the one-sample t test. (2)

To use a one-sample t-test, it is assumed that the numerical data are obtained independently and
represent a random sample from a population that follows a normal distribution.

In practice, it has been found that as long as the sample is not too small and the population is not

Experimental Design
Cria43
4
1
heavily skewed, the t-distribution provides an approximation to the sampling distribution of the mean

when σ (sigma) is not known.

4.29. Z test for the difference between two means. (2, 3, 7, 12 and various authors)

Z = (X1 - X2
)- (μ 1 - μ 2 )
/(σ 1
2
n1 )+ ( σ22 n)
1 2 1 1 2 2
Equation 16

X = Sample mean of population 1;

μ12 = Mean of population 1

σ12 = Variance of population 1; = sample size taken from population 1
n1

X = Sample mean of population 2;

2

μ2
2
= Mean of population 2
σ22 = Variance of population 2; = sample size taken from population 2
n2

The test statistic follows a standard normal distribution.

H o: μ1 = μ2 μ- μ= 0
1 2

H1 : #
μ1 μ2 μ - μ #0
1 2

Experimental Design
Cria43
4
2
4.30. t test for the difference between two means
The calculation of the t test for the difference between two means is done with:
t = (X1 - X2 )-(μ1 -μ2 ) / (Sp{(1 n1 )+(1/ n2 )}
Equation 17

S = ( - 1)S12 + ( -1)
n1 n2 S22 (n1-1) + ( - 1)
n2 Equation 18

S2p = pooled variance

X1,2 = sample means taken from the populations respectively
S1,2 = sample variances taken from the populations respectively

4.31. Degrees of freedom.

It is the number of excess observations taken to estimate a parameter. (Lipson and
Sheth 1973).

Or the number of variables minus the number of linear relationships (restriction).

4.32. Analysis of variance. (3, 7, 8, 27 and various authors)

The analysis of variance was devised by Sir Ronald Fisher in 1925, and is widely
used when an investigation aims to corroborate, through statistical analysis, the
effects of one or more factors on the behavior of a characteristic or dependent
variable.

As Anova (Analysis of variance) or Anava; is a statistical technique used to

analyze the total variation of the experimental results of a particular design,
breaking it down into independent sources of variation attributable to each of the
effects that constitute the experimental design, compares two or more means.

This technique aims to identify the importance of the different factors or treatments
under study and determine how they interact with each other.

The ANOVA is a test similar to the Student t test, in terms of practice, but the
comparison between groups is not through the mean and its standard deviation,
but through the variance of the numerical variable "y", in each group of the
categorical variable "x".

Basically, the analysis of variance is used to corroborate whether the significance

of differences between the means of two or more groups are or are not due to
chance.

The statistical figure obtained with the Anova is the F ratio.

Assuming that 2 groups are analyzed, the ANOVA analyzes the variations
between the two groups (inter-group) and compares it with the variation within
each group (intra-group), to obtain the F value through a sum of squares.

If the variance differences between each group are greater than those within
groups, there are surely significant differences between the groups that are not
due to chance.

Experimental Design
Cria43
4
3
Groups are defined as in the t test by choosing a categorical variable. The variable
to be analyzed must be numerical and have a symmetrical distribution.

4.33. Basics of variance analysis.

Suppose k independent random samples, of size n, drawn from a single normal
population. From them there are two independent ways to estimate the population
variance (S2)

1) A variance within groups (since only the variance within samples

contributes to it), or error variance, or mean squares of the error, and
usually represented by SCError (Mean Square Error) which is calculated as
the average of the k sample variances (each sample variance is a centered
estimator of S2 and the average of k centered estimators is also a centered
estimator and more efficient than all of them).

is a quotient: the numerator is called the sum of squares of the error and is
SCError
represented by SCE and the denominator is degrees of freedom because they are
the independent terms of the sum of squares.

2) Another called variance between groups (only the variance between the
different samples contributes to it), or variance of the treatments, or mean
squares of the treatments and represented by SCtrat or MSB (Mean Square
Between).

It is calculated from the variance of the sample means and is also a quotient; the
numerator is called the sum of squares of the treatments (represented by SCtrat)
and the denominator (k-1) degrees of freedom.

and SCError estimate the population variance under the hypothesis that the k
SCtrat
samples come from the same population.

The sampling distribution of the quotient of two independent estimates of the

variance of a normal population is a ratio with F distribution with the degrees of
freedom corresponding to the numerator and denominator respectively, therefore
this hypothesis can be tested using this distribution.

Total variation = Variation within treatments + Variation between treatments

Total sum of squares = Sum of squares within + Sum of squares between

Total variation = ε
Variation within them
treatments=
σ=.................................................................................................................................................
)- (μ1 - μ2 )....................................................................................................................................
)+(1/ n2 )}.........................................................................................................................................

Experimental Design
Cria43
4
4
 ANOVA GRAPH.....................................................................................................................
Example of decision making....................................................................................................
=∑ Yij......................................................................................................................................
=∑ ∑(Yij )2-[Y..2 N ]........................................................................................................................
=∑ ∑(Yij)2-[Y..2 N ].........................................................................................................................
H0 : μ1 =μ2 =μ3 =...μk =μ......................................................................................................
α1 = β2 = ω = ϑ....................................................................................................................................
μ1 = μ2 = μ3 = ...μk = μ.............................................................................................................................
Y = a+bX Equation 11..................................................................................................
a=∑Yb∑X n Equation 13...........................................................................................
r = {∑ XY -[∑ X ∑ Y n]}...................................................................................................
Equation 14.....................................................................................................................

Equation 20 -
X = mean of each group
Yo. IJ
-
X General = Grand average

These estimates are known as the sum of squares mean.

Once the estimates are obtained, the ratio between them is found, which is an F
estimate.

F= Estimate between treatments / Estimate within treatments.

If, based on this contrast, the hypothesis that SCtreatment and SCME have the same
variance is rejected, the hypothesis that the k means come from the same
population can be rejected.

Accepting that the samples come from populations with the same variance, this
rejection implies that the population means are different, so that a single contrast
tests the equality of k means.

There is a third way to estimate the population variance, although it is not

independent of the previous ones. If the kn (N) observations are considered as a
single sample, their sample variance is also a centered estimator of S2.

It is usually represented by CMT, it is called total variance or total mean squares, it

is also a quotient and the numerator is called the total sum of squares and is
represented by SCMT, and the denominator (kn -1) degrees of freedom.

The results of an analysis of variance (anova) are usually represented in a table

like the following:

Experimental Design
Cria43
4
5
 Table 4. Example of an analysis of variance

SC p-value
CM F0
Degrees of
Freedom Sum of Reason F
Source of Mean squares
squares
variation
Between Treatments k-1 SCtreatment SCtrat /(K-1) SCtrat/ SCME P (F>F0 )

Inside the Error Nk SCE SCE/(NK)

Total N-1 SCT

SCTotal/ (KN - 1)
K= Number of treatments or levels or groups of the factor of interest
N= Total number of data reported n = Total number of data per group

CMtratam = SCtratam/ (K-1) Equation 21

CME = SCE /(Nk) Equation 22

With F0 calculated as SCtrata/SCE Equation 23

And with a chosen α (generally 5%), with degrees of freedom in the numerator (k-
1), with degrees of freedom in the denominator (Nk) the upper critical value of the
F distribution is found, in a table of critical values like the example in numeral 4.8,
4.9.

The value of tabulated F (Ft) is found in the tables with 4 degrees of freedom in the
horizontal, that is, the numerator or treatments, and 15 degrees of freedom in the
vertical, that is, the denominator or degrees of freedom of the error. Furthermore,
with the significance level either 5% or 1%.

When comparing the calculated F value with the tabulated F value for 5 and 1%
significance, it is observed that Fc is greater than Ft, which indicates that there is
significance. Therefore, Ho is rejected, meaning that there are several treatments
that are better than the others.
If Fc is less than Ft the results are not significant. This indicates that the null
hypothesis that there are no differences between the treatments is accepted. The
significant value of F implies that the treatments do not belong to populations with
common μ. It indicates that the treatments differ significantly, but it does not tell us
or indicate which of them is better.
One of the bases for these tests is the standard error of the mean Sx and the
standard error for the difference between treatment means Sd. In the previous
example these values would be:
Sx = √ S2/r = √ 58.8/4 = √ 14.70 = 3.83 ;
Where S2 = mean square of the error or variance and, r = number of repetitions

Experimental Design
Cria43
4
6
 ANOVA GRAPH
Example of decision making

Figure 4. ANOVA decision making.

4.34. Least significant difference

It is the minimum difference that must exist between two sample means in order to
consider that two treatments are different.

When the hypothesis of equality between treatments is rejected for this design, the
researcher asks which of them are different from each other.
To find out, the methods of comparison of means are used, one of these methods
(formulas) is the LSD minimum significant difference (Equation 24) where the
number of replications is replaced by the number of blocks, and the degrees of
freedom of the error must also be changed in the case of blocks, it is given by (b-
1) (treatment-1).

For Ktreatments there is a total of k(k-1)/2 pairs of means.

K=treatments b=blocks N=Total experiences n=number of observations for each

treatment.
The value of tα/2, NK is taken from the Student T distribution tables with NK
degrees of freedom from error.

LSD = tα/2, NK a/2CME/n Equation 24

4.35. Experimental error

In every experimental process, two types of variations occur: the variation inherent
to the experimental material to which the treatments are applied and the variation
that comes from the lack of uniformity in the physical execution of the experiment
or variations external to the experiment (such as environmental variations, etc.)

Experimental Design
Cria43
4
7
that tend to mask the effect of the treatments.

To express these variations unrelated to the treatments, statisticians apply the

term experimental error; a term that does not mean mistake, but rather includes all
types of external variation unrelated to the experimental material.

This experimental error is the measure of variation that exists between

observations of experimental units in the same treatment, that is, the variation not
arising from the treatments.

For calculation purposes in the equations it is expressed with the letter ε

4.36. Type I and Type II errors. (8)
When contrasting or testing hypotheses, two types of errors can be made for the
following reasons:

• Hypothesis testing begins with an assumption, called a hypothesis, that we

make regarding a population parameter.

• We then collect and classify the data, apply the selected statistic to the
samples, and use this information to decide how likely it is that our
hypothesized population parameter is correct.

• We must establish the hypothesized value of the population parameter

before we begin taking the sample.

• The assumption we wish to test is known as the null hypothesis, and is

symbolized by H0.

• Whenever we reject the hypothesis, the conclusion that we do accept is

called the alternative hypothesis and is symbolized by H1.

• ** The rejection of a null hypothesis when it is true is called a type I error,

and its probability (which is also the level of significance) is symbolized as α.
***

• ** Accepting a null hypothesis when it is false is called a hypothesis error.

type II, and its probability is symbolized as β . ***

• The probability of making one type of error can be reduced only if we

wish to increase the probability of making the other type of error.

In order to obtain a low (probability) β, we will have to tolerate a high probability

α; however, the appropriate significance level must be decided according to the
costs and disadvantages associated with both types of errors.

Experimental Design
Cria43
4
8
4.37. The most recommended methods to reduce errors are:
In any experimental march that is carried out, it must be
a) Use very uniform experimental units, such as homogeneous raw materials,
processes, sample density, etc.
b) Adequate size of the experimental unit.
e) Elimination of competition between treatments.
d) Adequate distribution of treatments by drawing lots (chance).
e) Use the appropriate number of repetitions for each treatment.
f) Put all treatments on equal terms, so that if any is superior to the others, it can
be tested.

4.38. Statistical model. (Various authors)

The purpose of a series of experiments is to determine a statistical model that
reflects the belief regarding the relationship between treatments and observations.

Each outcome of the experiment, if repeated n times, is determined by the overall

mean μ and the treatment effect τ .

The identification of this phenomenon and the verification of the assumptions are
made at the time of proposing it based on the following mathematical model:

Yij = μ+ τi Mathematical model Equation 25

Where

i= 1,……., k number of treatments

j=1,…., r number of repetitions

μ = average of all experimental units in the experiment = X or Y as the case may

be.
τi = the difference between the average ( μ ) and the result when the combination
of factors determined by treatment i is obtained.

Since there are uncontrollable variations, the inaccuracy of the measurements

makes up a factor called experimental error ( ε ).

The model is finally expressed as:

Yij = μ τi +
+
εij Statistical model Equation 26

4.39. Analysis of Covariance.

What is sought in each experiment is the veracity of a hypothesis.

Experimental Design
Cria43
4
9
When the hypothesis refers to the parameters of the phenomenon, it then
becomes a statistical hypothesis such as the variability of the climate, of a
temperature in a process, even if measurement methods are available with
sufficient precision and accuracy, without forgetting that the result of every
experiment has a component called error, generally due to chance.

This random effect forces the researcher to resort to statistics to minimize it and
test the hypotheses with a significant degree of certainty.

Analysis of covariance is a very important procedure in experimentation for

verifying hypotheses, but unfortunately it is not used frequently.

The analysis of covariance allows adjustments to be made to the responses of the

experimental units in which a measurable factor is varied, with the response that
would have been obtained if all the experimental units had had the average value
of the variable factor, thus eliminating the effect of that factor.

The analysis of covariance among many utilities can also be used to adjust the
responses of the experimental units of averages of two or more variable factors
called covariates.

It uses analysis of variance and regression to eliminate the variability that exists in
the independent variable X; it also adjusts treatment means and thus estimates
much better the effect of the independent variable X on the dependent variable Y.

The independent variable X is an observation made on each experimental unit

before applying the treatments, and indicates to some degree the final response Y
of the experimental unit.

For example, in a farm assisted by UNAD, an experiment was carried out to study
the suitability of feed rations for weight gain in the above-mentioned cattle. For this
purpose, the initial weights X and feed consumption Y were determined. If the
researchers use different rations, the physical differences and effects presented by
the cattle may or may not be significant, due to the quality of the rations. Therefore,
the question must be asked: Since there is variation in the initial weights of the cattle,
is the difference in the final weights due to the properties of one of the rations? Or
that a high percentage is due to differences in initial weights?

Using covariance analysis, the effect of the quality of the rations can be calculated
by eliminating the part corresponding to the initial weights.

4.40. Recommendations
•When proposing a statistical study, clearly define the population being
analyzed.

Experimental Design
Cria43
5
0
 If you are working with samples, define the conditions they must meet
before extracting them.

• Specify what is to be measured, the units to be used, and how to record it.

4.41. Statistical software for problem solving.

Among the programs suitable for data management, processing and analysis we
mention the following:

Excel, Minitab, Statgraphics, SAS, Curve Expert; etc. However, with the use of a
calculator it is also possible to carry out the vast majority of operations involved in
data processing.

Figure 5. Cover of the Statgraphics program.

Experimental Design
Cria43
5
1
 Figure 6. Excel spreadsheet *.

* The name spreadsheet is assigned to a sheet that is divided into rows and columns,
the intersection of which is called cells, on which information (letters or numbers) is
stored that we can use to perform operations, such as addition, subtraction,
multiplication, division, financial, statistical, engineering, amortization calculations, etc.
4.42. Exercise developed
Practical exercise of study, reflection and analysis.

In a research project for the standardization of fruit, the following data were
found on the production of 50-kilo baskets of cape gooseberry of a previously
selected size on 50 farms in eastern Cundinamarca:

• ) Using the above data and applying the mathematical formulas provided,
check that the data obtained for this starting example in the fruit standardization
project can be presented as:

Table 5. Production of cape gooseberry baskets in six villages

Group 1 Group 2 Group 3 Group 4 Group 5

49 61 44 39 31
30 60 38 36 66
50 41 43 47 49
40 38 30 51 69
53 47 45 44 45
33 32 51 64 44
40 52 43 39 58
60 59 42 39 31
48 43 47 48 35
36 56 60 40 36

Experimental Design
Cria43
5
2
 Group 2, Group 5

Internal
Internal
variation
variation
2,….
5

Variation between groups

Figure 7. Variance between different cape gooseberry producing farms.

Table 6. Statistical summary of the total data collected on cape gooseberry

production in five municipalities of Cundinamarca.

Statistical characteristics List other properties of

found for the interest to consider in
standardization project this exercise.
50
Total samples (baskets)
Total sum 2282,0
Average 45,64
Variance 99,2963
Standard deviation 9,96475
30,0
Minimum value of cape
gooseberry production

Maximum value of cape 69,0

gooseberry production
Bias 1,33489
-0,66762
Kurtosis or sharpness

Experimental Design
Cria43
5
3
 Table 7. Statistical summary of data collected by month on cape gooseberry
production in five municipalities of Cundinamarca.

M No. of Average Variance Standard Standard Standard Summation by

on farms deviation Skewness Kurtosis group
th
1 5 44,8 126,2 11,2339 0,402954 0,1668 224,0
01
2 5 46,0 254,0 15,9374 0,482584 - 230,0
1,1925
8
3 5 46,0 15,0 3,87298 - - 230,0
0,392837 1,0041
6
4 5 45,6 227,3 15,0765 0,950036 0,3705 228,0
86
5 5 46,8 13,2 3,63318 1,62367 1,4570 234,0
1
6 5 44,8 177,7 13,3304 0,574125 - 224,0
0,3340
78
7 5 46,4 68,3 8,26438 0,693964 -86 232,0
8 5 46,2 163,70 , 12,7945 0,139123 - 231,0
7283 1,1074
7
9 5 44,2 30,7 5,54076 -1,4506 0,9761 221,0
77
1 5 45,6 132,8 11,5239 0,538284 - 228,0
0 1,3006

S 50 45,64 99,296 9,96475 1,33489 - 2282,0

u 3 0,6676
m 2
a- Review the following graphs and explain if any of interest are missing.

ANNUAL PRODUCTION OF CAPE GOOSEBERRIES

70 l 1

op
either
11
11
_ 60 1 1 11
AND
either Yo 1 1
1
Yo 50 1 1
50 1

• 1 1
• 1
1 i------ 1
< 40 1
--------
s 1
30 3------- 1 1
---------------
12345678910
MONTH

Figure 8. Chart of annual cape gooseberry production.

Experimental Design
Cria43
5
4
 Box and whisker chart

.................+.............. !
1
2 88888388888888888888888888898088888888888888888888888888888888888888 —
3 Yo ..............t.......... :
4 8838 88883833983903898988

EITHER 5 1 NE 1
6 ......................................................*...............................
7 888888888388888884000888888898088
8
9 1 ±............ 1
10 ...............................+...................................
—
30 40 50 60 70
CAPE GOOSEBERRY PRODUCTION

Figure 9. Box-and-whisker graph of monthly cape gooseberry production.

95% LSD MEDIA AND INTERVALS

FARMS

Figure 10. Graph of means and intervals for cape gooseberry

production per farm.

b- Also find the median, the mean, the absolute frequency, the number of
minimum samples.
c- With the data obtained we prepare the frequency distribution table.
d- Is the sample size taken correct?
d-What could be concluded and decided with the above data and what
suggestions would there be?
e- Complete the following Mariaza analysis table for the previous exercise with a
confidence level of 95.0%.

Table 8. Analysis of variance (ANAVA) for the example of cape gooseberry

producing farms.

SC p-value
Degrees of CM
Sum of F0 Reason
Source of Freedom. Mean squares
squares
variation

Experimental Design
Cria43
5
5
 SCtratam

Between groups 4 SCtratam /(k-1) 0,03 1

(farms)

Within groups = 45 SCE SCE /(Nk)

Error

Total 49 4865,52 SCMT / (kn -1)

• And also answer if this F-test value is greater than, less than, or equal to p
taken as 0.05 (5%)?

• Is there a statistically significant difference in cape gooseberry production

from one month to the next?

4.43. Exercises and activities

1 - Review an article from a journal with topics relevant to food engineering and/or
agricultural sciences:

a) Take the variables involved and classify them.

b) Identify the experimental design proposed by the researchers and exemplify the
concepts of indicator and dimension of a variable.
c) Do you consider the criteria given to decide whether the experiment was
successful to be appropriate?

2- An engineer conducts an experiment to compare the average absorption of

essential oil in a porous medium. To do this, he randomly selects a strip of five
possible absorbents and divides them into five pieces: one measuring 3.5
centimeters, another measuring five, the next measuring eight centimeters, the last
measuring ten centimeters, and the last measuring thirteen centimeters.
Identify the experimental unit, factors, and factor levels present in the experiment.

Could you attach another division and leave it to the environment? What would it
be called?

Could it be put another way?

3- In a supermarket operating under controlled conditions, the temperature
reached for 40 shelves (20 with tubers and 20 with fruits) was evaluated and the
following temperatures were reached:

Tubers 8,5 8,7 8,8 8,9 8,7 8,3 8,7 8,4 8,2
Fruit 8,4 8,2 8,7 8,6 8,3 8,5 8,3 8,5 8,4
a) Do you identify the treatments that were compared in this study?
b) Are the plant samples dependent or independent? Please explain.
e) Could we say that the average temperature of vegetables is the same for
tubers and vegetables?

4- The professionals of an experimental agricultural farm are trying to make a

decision for the purchase of a food supplement for which they evaluate the offers
of two suppliers of food rations by verifying the weight gain of two batches of cattle
and determined the following weight gains (Kgs).

Experimental Design
Cria43
5
6
 Supplier 1 Supplier 2
9 10
10 10
11 10
12 9
8 11
9 11

Calculate the mean, variance, standard deviation.

State the hypothesis and what would you conclude about the weight gain.
If we consider an α= 0.05.
B) If the company decides to try another supplier, what criteria would you apply?
The following information was obtained for this plugin:

Supplier 3
8
8
6
12
12
13

Objectively test whether there is a significant difference between suppliers. Work to

be done at home and handed in to the tutor.

UNIT TWO: Experimental designs in food engineering

Chapter Five: Classification of experimental designs
5.1. Characteristics of experimental designs. (Various authors)
Experimental designs must have, among others, the following characteristics:

a. Simplicity. The selection of treatments and the experimental layout should be

made as simple as possible.
b. Degree of precision. The experiment must have the capacity to measure
differences between treatments with the degrees of precision desired by the
researcher. To achieve this goal, you must start with an appropriate design and
number of repetitions.
c. Absence of systematic error. An experiment should be planned with the
purpose of ensuring that experimental units receiving one treatment do not differ
systematically from those receiving another treatment, thus attempting to obtain an
unbiased estimate of the treatment effect.
d. Range of validity of the conclusions. This should be as broad as possible.
Experiments that contribute to increasing the range of validity of the experiment
are replicated experiments and experiments with factorial structures.
e. Calculation of the degree of uncertainty. In every experiment there is some
degree of uncertainty regarding the validation of the conclusions. The experiment
must be designed so that it is possible to calculate the probability of obtaining the
observed results due solely to chance.

Experimental Design
Cria43
5
7
5.2. Influential aspects in the selection of an experimental design. (Various
authors)
It can be said that the number of experimental designs is as wide as the number of
situations to be studied and responded to, given the great diversity of problems or
situations that occur in the daily practice of the professions.

Hence, it is important to know how to choose the most appropriate one for the
situation or problem you want to solve, therefore the classification of designs
according to their objective and their scope within the objective in question.

The five aspects that most influence the selection of an experimental design are:

a. The objective of the experiment.

b. The number of factors to control.
c. The number of levels that are tested in each factor.
d. The effects of interest to investigate (factor-response relationship).
e. The cost of the experiment, time and desired precision.

5.3. Classification of experimental designs according to their use.

(Various authors)

In order to solve the large number of theoretical and practical situations and
problems that arise in engineering and other professional activities, a good number
of experimental designs have been proposed, making it necessary to know how to
choose the most appropriate one for the situation to be solved. Therefore, it is
convenient to know their classification according to their scope within the proposed
objective.
In this sense, according to their objective and without claiming to be exhaustive,
designs can be classified as:

1. Designs to compare two or more treatments.

2. Designs to study the effect of several factors on the response(s).
3. Designs to determine the optimal operating point of a process.
4. Designs for the optimization of a mixture.
5. Designs to make the product insensitive to uncontrollable factors.

Examples of the above are those that will be discussed in this course, such as
those mentioned below:

Complete Randomized Design, Complete Randomized Block Design, Latin Square

design, Greco-Latin Square design, 2K factorial design, 3K factorial designs.

5.4. One variable experiments

One-variable experiments are done to find out how one experimental variable
affects one or more response variables. In this type of experiment, treatments are
simply selected levels of the experimental variable.

Depending on the experimental variable selected, treatments may differ either

qualitatively or quantitatively.

5.5. Designed unblocked experiments. (Various authors)

Some authors call these experiments totally randomized or completely random

Experimental Design
Cria43
5
8
designs. It is the simplest design and is used when the experimental units are
homogeneous, have the same chances of receiving any treatment and the
variation between them is very small.

As in laboratory experiments, greenhouse experiments where environmental

conditions are controlled, in some cases this design requires only one replicate
with a single classification criterion.

Therefore:

1. It's easy to plan.

2. It is flexible in terms of the number of treatments and repetitions, the limit
is given by the number of experimental units in general.
3. The number of treatments does not need to be equal to the number of
repetitions.
4. No lost units or samples are estimated.

5. The number of degrees of freedom for error increases by not having many
restrictions.

The disadvantages of the completely randomized design are:

1. It is not efficient with heterogeneous experimental material.

6. Since there are no restrictions on randomness, the experimental error
includes the total variation between experimental units.

The analysis of variance for the completely randomized design is shown in Tables
25 and 26.

5.6. Statistical procedure for testing hypotheses in a randomized design.

The procedure to follow is common and must be in accordance with what is set out
in the following table.

Table 9. Statistical procedure for hypothesis testing

INFORMATION COLLECTED ON THE RESPONSE

TO TREATMENTS OPERATIONS
ARITHMETICS TO BE
DEVELOPED FOR THE
CALCULATION OF ANAVA
TO B C 1. = sum of
squares of
F1A1 F1B1 F1C1 all the
OBSERVATIONS F2A1 F2B1 F2C1 Observations
2. Sum of the
data
F3A1 F3B1 F3C1
3. Total number of
Total by measurements
Yi.1 4. Data mean
treatment

Experimental Design
Cria43
5
9
Number of data per n N n
treatment
Average by n

treatment
Yi=
j=1
∑ Y
5. Effect of
the
Yi. = Yi. n method
Big media AND.. = And.. N N = an 6. And observed - And a
half

Addition of SCT
total squares
Addition of SCt
squares of the
treatments

Experimental Design
Cria43
6
0
 And the ANAVA in the following table

Table 10. Analysis of variance for treatments with a single factor, in a completely
randomized design.

Degrees of SC F0 Reason p-value

Freedom CM
Sum of
Mean squares
squares
Source of variation
Treatments k-1 SCtrat SCtrat /(k-1) SCtrat / CME P (F>F0 )

Mistake Nk SCE SCE /(Nk)

Total N-1 SCT SCMT/ (kn -1)

k= Number of treatments or levels or groups of the factor of interest
N= Total number of data reported
n = Total number per group

The procedure consists of calculating the sum of the squares of the treatments,
the sum of the total squares and by difference the sum of squares of the error.

Table 11. collection and presentation of treatment results, in a completely

randomized design.

LEVE 1 2 3
L
TO F1A1 F2A1 F3A1 ∑ ∑
Yi.1 X1 Yi..1X
B F1B1 F2B1 F3B1

C F1C1 F2C1 F3C1

D F1D1 F2D1 F3D1

AN F1A1 F2A1 FjeAJ

Yi.1
D
∑ CT ∑ X
F1A1 to FJEAJ are the analyses or samples collected or taken

Yij is each of the data represents two as

F 1 A 1 to Fje

Experimental Design
Cria43
6
1
∑ X = Y.. = Grand
n
Yi. ==1∑Yijj Equation 27 Yi.
average
Equation 28
Y..=∑ ∑Yij Equation 29 j=1
a=1
And..= N Equation 30
N = an Equation 31 And..

N = total data recorded

a = number of levels of = A + B + C + D + E

n = number of repetitions = 1 + 2 + 3 + etc.

a n- ∑⎜⎛Yij-Y..
SC .
=∑
i=1
j=1 ⎝
Total
Equation 32

∑
- -2 n -2
to to

⎜⎛Yi.-Y.. ⎟ ⎞
∑⎜⎛Yij-
SCTotal
=
n Yi.⎟⎞j
Equation 33

∑
i=1
=1 ⎝ ⎠
+ ⎝ ⎠ i=1

∑ ∑(
an

SCTotal = Yij 2-[

) i =1 j=1 Equation 34
n
H. treatments
H = ∑ Yij
n-
[Y.. /N ]
2
Equation 35
i=1

SSTotal = SStratam + SSerror Equation 36

Mean Square of the trat = CMtratam = SCtratam /(k-1) Equation 37

Mean Square Error = CME= SCerror /(Nk) Equation 38

With these calculate the F0 statistic as CMean of treatmentT / CME Equation 39

And with a chosen α (generally 5%), with some degrees of freedom in the
numerator (k-1) of the treatments, with some degrees of freedom in the
denominator (Nk) of the error, the upper critical value of the F distribution is found,
in a table of critical values like the one in the appendices.
The comparison P (F>F0 ) is performed to determine the proposed hypotheses.

This design follows a statistical model. (24)

Experimental Design
Cria43
6
2
5.7. Statistical model for a randomized design. (Various authors) It is generally
represented by the following equation:

Yij = μ+τi + εij μ= global mean τ i = treatments

εij = error
Equation 40
5.8. Example of an unblocked
experimental design
The transportability, handling and presentation properties of the doughnuts
produced by a bakery are to be improved. To this end, the effect on the hardness
of the doughnuts of the addition of three additives in the same proportion will be
determined; the additives will be incorporated individually into the dough produced
(knead).

5 6

Ingredient A Ingredient C

Figure for the production of doughnuts from a

standardized mass. 12.
Flowchart

For the study, an unblocked experimental design will be used.

• Research hypothesis: The hypothesis is that at least one of the three

ingredients would improve the final hardness of the donuts.

Experimental Design
Cria43
6
3
 • Experiment design: a homogeneous dough for making donuts is divided
into six equal portions and the selected ingredient is mixed into each two
of them.

Hardness is evaluated after six days of storage at 10*C.

The first step is to randomize by assigning a sequence of random numbers to the

experimental units, using one of the permutation tables from which a random
permutation is taken, for example:

318594267, assign the first treatment to the two portions of treatment A, the next
two to treatment B and finally the remaining two to treatment C

The randomization performed is summarized in the following table.

Table 12. From randomization to the donut example

Portion of
dough 3 1 7 5 4 8 2 6 9
Treatment or
additive TO TO TO B B B C C C

Solution: with the data in the table, perform the indicated operations n
Yi.=∑Yij=7+9 +5+10+8+9+10+9+12 = 79 Yi.=79/9=8.77
j=1

∑ ∑(
an

Mr
Total = Yij )2-[Y.. N 2
]
i =1 j =1

SCT= {(7) +(9) + (5) +(10) + (8) + (9) + (10) +(9) + (12) }-
2 2 2 2 2 2 2 2 2
{ (8.77) /3x3 }= …
2

SCtreatments= [{(21) +(27) + (31) }]/3 -- {(8.77) /3x3 } = ……..

2 2 2 2

SCT=SCE+SC SCT– SCt = SCE SCE =

t

Experimental Design
Cria43
6
4
 Table 13. Data obtained in the determination of hardness for the example of
donuts.

Treatments Hardness or response variable

Determination Total Averag Variance
(Und/cmt) e
3 7 1
Additive A 7 9 5 21 7 2

5 8 4

Additive B 10 8 9 27 9 1

2 9 6

Additive C 10 9 12 31 10.33 1.527

Big media 8.77

Table 14. Analysis of variance to evaluate the effect of additives in the production
of donuts.

Degrees of SC p-value
CM F0 Reason
Freedom Sum of
Source of Mean squares Equation
Ecuac squares
variation Ecuac
Ecuac
3-1=2 8,44444 3,45
16,8889 0,1004
Treatments
Mistake 6 14,6667 2,44444
Total (9-1)=8 31,5556 SSTotals/ (kn -1)

If we also use the multiple range test for donuts: 95% LSD

Table 15. Homogeneity and contrast analysis of the donuts.

Treatment Samples AverageGroup Contrast (Difference)

homogeneity
TO 3 7 X AB ( -2)
B 3 9 XX BC (-3.33)*
C 3 10.33 X BC (-1,33)
* Indicates statistically significant difference
Checking that the selected treatment is the one that presents the best conditions
to be applied to the donut.

Chapter Six: Block Design

6.1. Randomized complete block design. (8)

Experimental Design
Cria43
6
5
 6.1.1. Characteristics.
The objective of the randomized complete block design (RCD) is to gather the
experimental units to which the treatments will be applied, in blocks of a certain
size, in such a way that the treatments are carried out within each block, the
variability between experimental units of different blocks will be greater than
between units of the same block, as a consequence, the differences found
between units are mainly due to the discrepancy between treatments. The
disparity that is not due to treatment is eliminated by design and forms part of the
experimental error. According to this, it is easy to observe that the variability
between blocks does not affect the differences between treatment means,
because in each block it appears once per treatment and thus the blocks and
treatments are orthogonal.

In this design the experimental material is divided into groups of experimental

units (EU) that are as homogeneous as possible. Each group is a single test or
repetition.

The object at all stages of the experiment is to keep the experimental error within
each group as small as practically possible.

Sets or groups are called blocks.

Within each block the (EU) are randomly assigned, each treatment occurs
exactly once in a block.

6.1.2. Sources of variability.

In this design (DBCA) three sources of variability are considered: the treatment
factor, the block factor and the random error, that is, there are three possible
causes of the variability presented by the data obtained.

They are called complete because all treatments are tested in each block, that is,
the blocks are complete.

6.1.3. Advantages
The main advantages of random blocks are as follows:

1. Clustering often produces more accurate results than completely

randomized designs. Coch.
2. Any number of repetitions may be included.
The statistical analysis is the usual one.

Experimental Design
Cria43
6
6
 3. If the variance of the experimental error is greater for some treatments
than for others, an unbiased error can still be obtained for testing any
specific combination of treatment means.
4. No other design is used as frequently as random blocks.

6.1.4. Randomization.
Randomization is performed when units have been grouped together and
treatments have been randomly assigned to units within each group.

The name of complete randomized blocks is due to the fact that in each block
(for example an operator, a cattle, a plot) all the treatments are tested (complete)
in random order.

6.2. Statistical procedure for testing hypotheses in a randomized

complete block design.
The procedure to follow is common and must be in accordance with what is set
out in the following table.

Table 16. Collection and presentation of treatment results, in a randomized

complete block design.

BLOCKS
Treat B1 B2 B3 B4 Yi.
r1 XT1B1 XT1B2 XT1B3 XT1B4
X
R1
r2 XT2B1 XT2B2 XT2B3 XT2B4
X
R2
r3 XT3B1 XT3B2 XT3B3 XT3B4

r4 XT4B1 XT4B2 XT4B3 XT4B4

∑XR3
Yj
Yi. B1 Yi. B2 Yi. B3 Yi. B4 ∑XR4
∑ X Btotals = ∑
∑X=Y..

bb

∑ X = Y.. = Grand
average
Yi. =
j=1
∑ Yij Yi. =
j=1
∑ Yij Equation 41
to

Ys = ∑ Yij Equation 42
j =1

Equation 43 Yij=Yj to Equation 44

Experimental Design
Cria43
6
7
AND.. = And.. N Grand mean Equation 45
N = total recorded data
N = an a = number of levels = A + B + C + D + E
n = number of repetitions = 1 + 2 + 3 + etc.
ab to
Y..=∑∑Yij=∑Yi=∑Yj
=1 j=1
i =1
i
Equation 46

∑∑
n

SCTotal = (Yij)2-(Y..) N 2
Equation 47
j=1
to

H.H.. =[ ∑(Yj) 2
]-[(Y..) 2
N] Equation 48
Blocks i=1

]- Y..) ]
to
Yi)
SStreatments = [ ∑( 2
b [( 2
N Equation 49
i =1

SSTotal = SSBloques + SSError Equation 50

6.3. Arranging data in a DBCA design. (Various authors) It is customary to follow

the following arrangement.

Table 17. Arrangement of the data in a randomized complete block design.

TREATMENTS
1 2 3 4 … J YiB
1
2
And
Y12
... Y22
Y21
Y31
Y32
Y41
Y42
…
….
Yj1
Yj2 Yi. B1

BLOCKS
3
4
Y13
Y14
Y23
Y24
Y33
Y34
Y43
Y44
…
..
Yj3
Yj4 Yi. B2

Yi. B3
.AND …Yjf
. ........1f Y2f Y3f Y4f ..
.
F
Yi. Bn.
Y.JBJ Y.JB1 Y.JB2 Y.JB3 Y.JB4 Y.JB.. Y.JBN

6.4. Statistical model for a DBCA

Any observation in the DBCA can be represented by the model;

Yij = μ + τi + γ j + εij Equation 51 i= 1,……., k number of treatments j=1,….,

b number of blocks

γ j = effect due to the block

6.5. Hypothesis to be tested

The hypothesis tests that can be carried out depend on the type of model used.

Experimental Design
Cria43
6
8
 • These conclusions may be subject to error:
• Type I error: rejecting the equality of the treatment means when they do
not differ from each other.
• Type II error: not rejecting the equality of the treatment means when they
differ from each other.

H0 : μ1 = μ2 = μ3 = ...μk = μ Equation 52

Ha: μ #μ 1 j for some i # j

6.6. ANOVA for a randomized complete block design (RBCD) (Various

authors)
The analysis methodology to be used is ANOVA with two classification criteria. In
this case, the analysis of variance partitions the total variability of the information
into three components: the first due to the effect of the treatments (Sum of
Squares Between Treatments), the second due to the effect of the blocks (Sum of
Squares of Blocks), and finally the Experimental Error (Sum of Squares of the
Error).

The degrees of freedom are partitioned in the same way.

Whether or not the null hypothesis is rejected depends on the value of the F
statistic:

F = (S. C. OF TREATMENTS / T - 1) / (S. C. OF THE ERROR / (T - 1) (R - 1)) EQUATION 53

Table 18. ANOVA for a randomized complete block design

Sum of Degrees of Mean Fo p-value

squares freedom Squares
Ecuac Ecuac Equation Ecuac

Treatments SCtrat k-1 SCtrat P(F > Fo)

Fo = SCtrat / CME

Blocks SCB b-1 CMB Fo = CMB / P(F > Fo)

CME

Mistake SCE (k -l)(b -1) CME

Total SCT N-1

Experimental Design
Cria43
6
9
6.7. Comparison of treatment means in the randomized complete block
design (BCA)
When the hypothesis of equality between treatments is rejected for this design, the
researcher asks which of them are different from each other.
To find out, the methods of comparison of means are used, one of these methods
(formulas) is the LSD minimum significant difference (Equation ….) where the
number of replications is replaced by the number of blocks, and the degrees of
freedom of the error must also be changed in the case of blocks, it is given by (b-
1)(tratam-1).

For K treatments there is a total of k(k-1)/2 pairs of means.

K=treatments b=blocks N=Total experiences n=number of observations for each

treatment.
The value of tα/2, NK is taken from the Student T distribution tables with NK
degrees of freedom from error.

LSD = tα/2, NK -J2CME / n Equation 54

6.8. Example of randomized complete blocks (RCB) (Various authors)

A juice processing company is investigating the effect of four evaporators on the
evaporation rate (concentration) of tomato juice from four locations across the
country.

For this purpose, juice is made from tomatoes from the same location and each
one is concentrated in four shifts in the four evaporators in a completely random
order.
The time taken to reach a concentration of 20 degrees Brix is determined as a
response variable.
The results of the 16 tests are presented in Table 9.

Please perform the Indian operations and discuss them with the tutor…..

Table 19. Data obtained in the determination of solids content for the example of
tomato juice concentration.

Treatments = concentration methods

Experimental Design
Cria43
7
0
 1 2 3 4

BLOCKS = 1 1.7 2.3 1.8 2.1

2 2.3 2.08 2.2 2.4
Shifts 3 2 1.9 2.1 1.95
4 1.8 2 2.3 2

To make this table, the procedure that has been followed through the last five
numerals is followed.

Table 20. Processing of data obtained in the determination of solids content for the
example of tomato juice concentration.

Methods of concentration Socks

Totals per block
1 2 3 4
1 1.7 2.3 1.8 2.1 7,9 1,975

Shifts 2 2.3 2.08 2.2 2.4 8,88 2,245

3 2 1.9 2.1 1.95 7,95 1,9875

4 1,8 2 2.3 2 8,1 2,025

7,8 8,28 8,4 8,44 33.83

Totals by treatment
Average per 1,95 2,07 2,1 2,11
treatment

∑
n

Yi. = Yij = 1.7+2.3+1.8+2.1+2.3+2.08+2.2+2.4+2+1.9+2.1+1.95+….

+2.3+2=……… j=1
Yi.=79/9=8.77

∑ ∑(
an

SC .
Total
= Yij)2-[Y.. N 2
]
i =1 j =1

SCT= {(1.7)2+(2.32 + (1.8)2+(2.1)2 +….+ (2)2+ (2.3)2+(2)2 }- {(33.83)2/4x4 }= …

SCtreatments= [{(7.8) +(8.28) + (8.4) +(8.44)}]/4 -- {(33.83) /4x4 } = ……..

2 2 2 2

SCBLOCKS = [{(7.9) +(8.88) + (7.95) +(8.1)}]/4 -- {(33.83) /4x4 } = ……..

2 2 2 2

Experimental Design
Cria43
7
1
 SCT=SCE+SC SCT– SCt = SCE SCE= ….
t

And the ANAVA developed for this example is the following:

Table 21. ANAVA for the data obtained in the determination of the solids content
for the example of tomato juice concentration.

p-value
Sum of Degrees of Mean Fo
squares freedom Squares
Ecuac Ecuac Equation Ecuac

MET DE 0,0661688 3 0,0220563 ?? 0,6662

CONCENTRA
Treatments

Blocks B: 0,191669 3 ?? 1,57 0,2641

OPERATORS

Mistake 0,367006 9 0,0407785

Total 0,624844 15

?? You must complete the missing pieces in table 9 above.

None of the factors has a significant effect on the time of concentrations at the
95% confidence level.

6.9. Exercise to do as a group activity:

Write an essay on the following question:

When the number of experimental units is not equal, what procedure is

performed?

6.10. Proposed exercise

The professionals of a cheese producing company are carrying out a study on the
way of presenting and transporting a variety of cheese (Q) that they market, for
which they have carried out a series of measurements on the hardness (of the
same) in six samples from different processes using three measuring instruments
(P) for this purpose...
Obtaining the following results:

P1 P2 P3

Experimental Design
Cria43
7
2
 Q1 9,5 12,3 12
Q2 10 13 10,8
Q3 11,4 11,2 10,6
Q4 11,2 9,6 10,2
Q5 9,8 11 10,2
Q6 9,4 9,1 10,3

-Test whether the differences between the hardness of the cheeses are significant
and whether the difference between the measuring instruments is significant.
Raise the respective hypotheses, and what would you conclude about the weight
gain.
If we consider an α= 0.01.
Chapter Seven: Latin Square Design
7.1. Latin square design
The name of Latin square is due to RA Fisher [The Arrangement of Field
Experiments, J. Ministry Agric., 33: 503-513 (1926)]. The first applications were in
the agricultural field, especially in cases of soils with fertility trends in two
directions.

Design that contemplates the control of two block factors, in addition to the
treatment factor, and the three factors have the same number of levels.

Considering four sources of variability, which are the two block factors, the
treatment and the randomized error.

The first blocking factor is represented in the rows, the second blocking factor is
represented in the columns, and the treatments are represented by Latin letters
and are distributed in such a way that each treatment appears only once in each
row and only once in each column (Latin square design). The interest is focused
on a single factor, the treatments, but two restrictions are imposed on the
randomization in a table such as the example in the table in the following section.

7.2. Formation and tabulation of experimental data in a Latin square.

Suppose 4 treatments A, B, C and D, with these treatments 4 different tables
called typical or standard can be formed (in the first row and in the first column
there is the same distribution).

Experimental Design
Cria43
7
3
 Table 22. Compilation and presentation of treatment results, in Latin type.

TO B C D TO B C D TO B C D TO B C D
B TO D C B C D TO B D TO C B TO D C
C D B TO C D TO B C TO D B C D TO B
D C TO B D TO B C D C B TO D C B TO
From each square 144 different shapes are obtained, in total there are 576 different
squares.

Such a design is only possible when the number of levels of both restrictions is
equal to the number of levels of the treatment.

For example, in an experimental farm if we administer three drugs (treatments) to

three goats with different . The order in which the goats receive the treatments can
be completely randomized (block design) or randomized under the “balance”
condition required for a Latin square. Let us designate the treatments by and . T1, T2 T3

A balanced allocation with respect to the order of administration can be

7.3. Arrangement of data in a Latin Square design.

To facilitate the understanding of this design, the example recommended by
several authors and presented in the following table should be followed:

Table 23. Arrangement of treatment results, in Latin table.

BLOCKS II
1 2 3 4 … J Yi..

1 B=Y221 C=Y331 D=Y4411 Yi..1

A=Y11 F=YjJ1
2 C=Y322 D=Y432 E=Y442 Yi..2
B=Y212 A=Y1j2
3 C=Y313
D=Y423 A=Y533 B=Y443 B=Y2j3 Yi..3
BLOCKS I 4 D=Y414
F=Y524 F=Y634 A=Y544 .. C=Y3j4 Yi..4
. Yi..,,
. F=..Y... f1f J=…YJjf
F A=Y12f B=Y23f C=Y34f Yi..f

Yj.. Y1.. Y2.. Yj.. Yj.. Ynj.. Yj.. Y...= ∑ Yijk

The choice of the Latin box is made before obtaining the data.
7.4. Statistical procedure for testing hypotheses in a Latin square block
design.
The procedure to follow is common and must be in accordance with what is set out
in the following table.

Experimental Design
Cria43
7
4
 And... = the sum of all the data in the results table

Y...= ∑ Yijk
Equation 55

Yi.. = Equation 56 And j.. = Yj,\..an Equation 57

Yi..
Yippee. n Equation 58
= And...

SCTotal = ∑ ∑ ∑( Yijk )2-[Y... N

2
] Equation 59 i jk

= ∑ [
t

SStratamien tto Yij2k Equation 60

a ]-
And...
j =1
Note: Treatments are defined with Latin letters

[
b2

SSrows = ∑ Yij2k b ]- And...

N Equation 61
i =1

= ∑ [
t

SScolumns Yij2k Equation 62

c ]-
And...
k =1

*Rows first blocking factor

**Columns second blocking factor

A=b=c
The degrees of freedom are cal

SSTotals= SSrown+SScolumn+SStratamien+SSError Equation 63

7.5. Analysis of the design of the Latin Square (Various authors)

Any observation in the DBCA can be represented by the model;

Y, — LI—T, “1“ S, —8,, - ■■

ijl μ i γ j l ijl Equation 64

i= 1,……., k number of treatments

j=1,…., b number of blocks

γ j = effect due to block = effect due to treatment

δi 1.19.2 Hypothesis to be tested

Experimental Design
Cria43
7
5
 H0 : μ1 = =
μ2 μ3 =... =μ μk

Ha: μ1 # μj for some i # j

7.6. ANAVA for the Latin frame design.

The variance analysis is performed following the traditional method to compare
three factors and the same number of factors. To this end, the boxes in Table 36
must be filled in and, as a practical example, in Table 23.

Table 24. ANOVA for Latin square design*

SOURCES OF SUM OF DEGREES MEDIUM

Fo P-value
VARIABILITY SQUARES OF SQUARES
FREEDOM

Treatments SCTRAT k-1 CMTRAT P (F > Fa)

F = CMTRAT/
CME

Lines SCBL k-1 CMBl F = CMB l P(F > Fa)

/CME

Columns SCB2 k-1 CMB2 F= P(F > Fa)

CMB2/CME

Mistake SCE (k - 2)(k -1) CME

Total SCT k2
-1

*Taken from Gutierrez, H, et all. Analysis of experiments. Chapter Four 132

SCT = SCtrat + SCBl + SCB2 + SCE Equation 65

and the corresponding degrees of freedom are

k2
-1 = (k -1) + (k -1) + (k -1) + (k - 2)(k -1) Equation 66

7.7. Example for the design of a Latin square.

In a food processing industry, it is decided to check whether the weight loss
determined in grams of a product stored in four different packages A, B, C and D
is significant. For this purpose, four different technologists I, II, III and IV are asked
to verify it, using four different scales 1, 2, 3 and 4 for this purpose. The
determinations are carried out following the Latin square model in the following
table:

Experimental Design
Cria43
7
6
The results of the 16 tests are presented in Table 25.

Table 25. Data collected by technologists

Weighing Methods - Scales
Technologist 1 2 3 4
Yo A= 503 B= 507 C=510 D=512

II B=514 A=493 D=495 C=496

III C=507 D=509 A=509 B=505

IV D=511 C=517 B=499 A=520

Procedure

1- Calculate the SCtreatments (sum of squares) according to the formula

… and table …….
1.1. For which the sum of all treatments A, B, C and D must be
calculated as:
=503+493+509+520=2025
∑A
=512+495+509+511=2027
∑D

Each of these sums is replaced in the treatment equation.

2- Calculate the row SC (sum of squares) according to the given
formula.
3- Calculate the SCcolumns (sum of squares) according to the formula

4- Calculate the SCTotales (sum of squares) according to formula Ec 10.

5- Determine the error by difference.
SCE= SCT - SCtrat - SCBl - SCB2
6- Determine the corresponding degrees of freedom Equation ….
7- Calculate the ANAVA table…….
8- What can be concluded?

Experimental Design
Cria43
7
7
 The student must select the appropriate formula from those previously
proposed.

Table 26. Calculation of treatments (3rd or letters)

Latin letter Total treatment In this example

TO A= ∑ Yl1l A = 2025

B B=∑ Y221 B =2025

C C =2030
C=∑ Y331

D D =2027
D = ∑ yi3

Table 27. ANOVA for Latin Square Design of Weight Loss Example

SUM OF DEGREES OF MEAN

SOURCES OF
SQUARES FREEDOM SQUARES Fo Ecuac P-value
VARIABILITY
Ecuac Ecuac Equation

Weight-balance 74,1875 3 24,7292 0,28 0,8355

determination

Operator ?? 3 106,229 1,22 ??

Packing 4,1875 3 ?? 0,02 0,9969

Mistake 522,375 6 87,0625

Total 919,438 15

?? You must complete the missing pieces in table 27 above.

There are no P values that are less than 0.05, and no factor has a
statistically significant effect on weight loss at the 95% confidence level.

What would be your conclusion and what could be recommended for

exercise?

Experimental Design
Cria43
7
8
 7.8. Proposed exercises

1) In order to determine whether there are effects on weight due to the

position that broiler chickens can occupy, raised in a cage with 4 floors
(rows) and 4 boxes (columns). The variable analyzed was: Chicken
weight (kg.) at 8 weeks of age

Lockers
Flats 1 2 3 4
1 1.40(A) 1.38(B) 1.40(C) 1.60(D)
2 1.35(B) 1.28(A) 1.45(D) 1.62(C)
3 1.38(C) 1.40(D) 1.42(B) 1.63(A)
4 1.39(D) 1.39(C) 1.40(A) 1.60(B)

a) Formulate and perform the corresponding Hypothesis Test. Use =0.05 α

b) Perform the Tukey test to compare whether there is a difference between
treatment A and B. Use =0.05 α
c) Perform DLS test to compare if there is a difference between treatment C
and D. Use =0.01 α
d) Use the t-test to compare whether the average weight using treatment C
is less than the average weight using treatment B. Use α=0.05
Chapter Eight: Design in Greco-Latin painting
8.1. Design in Greco-Latin style.
When there is a fourth block factor to be actively controlled in the Latin table
design, such as if in the exercise of numeral 1.19.5 it is decided to consider
the site of the experiment or the origin of an ingredient for the product
considered.

This design is called a Greco-Latin table (GCLT), in which the levels of the

new factor are denoted by the Greek letters , α β, δ, γa; the treatments are
represented by the Latin letters A, B, C and D.

In the Greco-Latin square design, two Latin squares are superimposed,

resulting in the following mathematical model:

Yijl = μ+ τi +γj +δl + βk + εijl Equation 67

i= 1,……., the number of treatments

j=1,…., b number of blocks
τi = effect due to block 1

Experimental Design
Cria43
7
9
γj = effect due to block 2
δl = effect due to treatment
ηl = cto due to treatment2 { αl , β,ϑ,ω }
The drawback of this model is that its use is very restrictive. Furthermore, Greco-
Latin squares may not exist under certain conditions.

8.2. Formation and tabulation of experimental data in a Greco-Latin square.

Table 28. Arrangement of treatment results, in a Greco-Latin table.

BLOCKS II AND ..k

1 2 3 4 Yi…
2 αA=Yl1l
F=
βB=Y22 ϑC=Y331 ωD=Y4411 YjJ1
BLOCKS I 2 A=Y1j2
And i..
4 C=Y322 αA =Y C=Y3j4
3. βB=Y212 ωD=Y42
ϑC=Y313 ϑ ωD=Y432533 αA=Y544 βB =Y443 B=…Y2j3
3j4

. ωD=Y414 αA=Y12f βB=Y23f ϑC=Y34f …

F J=YJjf

And…l

8.3. Statistical procedure for testing hypotheses in a Greco-Latin block

design.
The procedure to follow is common and must be in accordance with what is set out
in the following table.

AND = the sum of all the data in the results table

AND = ∑ Y ijkl

∑ ∑ ∑ ∑( ]
an

Equation
SCTotal = Yijkl 2-[ Y... N
) 2
ijk l1
68

letter treatment
t

SLatin Equation 69
= ∑[ Y. .
2j

j =1

Note: Treatments are defined with Latin letters

Experimental Design
Cria43
8
0
SSrows = ∑ [ 2
Yi. .. p ]- And 2
i =1 N Equation 70

= ∑ [ ]-
Pt
2

SScolumns Y p
..2..L And N Equation 71
L =1

SStratamletrgriega = ∑ [Y.. 2k. p

=1
2

k
] -
AND N Equation 72

*Rows first blocking factor

**Columns second blocking factor Treatment
represented by the Greek letters

α 1 = β 2 = ω=ϑ
The degrees of freedom are cal

SSTotals=SSrown+SScolumn+SStratamienLatin+SStratamien;etrgriega+SSError Equation 73

8.4. Analysis of the design of a Greco-Latin painting (Various authors)

Any observation in the DBCA can be represented by the model;

Yijl = μ +τi +γj +δl + εijl Equation 74

i= 1,……., k number of treatments

j=1,…., b number of blocks
= effect due to the block
γj
δi = effect due to treatment

8.5. Hypothesis to be tested for Greco-Latin design

H0 :
μ1 = μ2 = μ3 = ... = μk

μ
Ha: μ1 # μj for some i # j

Experimental Design
Cria43
8
1
 8.6. ANAVA for Greco-Latin design.

Table 29. ANOVA for the Greco-Latin square design*

SOURCES OF SUM OF DEGREES MEDIUM

Fo P-value
VARIABILITY SQUARES OF SQUARES
FREEDOM

Treatments Latin SCtratlatin k-1 CMtratalatin P (F > Fa)

letter Ftralat = CMtratlatin/
CMEr

Lines SCrenglo k-1 CMrenglo P(F > Fa)

Fre = CMrenglol
/CMEr

Columns SCcolumn k-1 CMcolumn F colmn= P(F > Fa)

CMcolumn/CM
It is
Greek Letter SCtratgiega k-1 CMtaratGreek Ftratlgrieg=CM
Treatments taratgriega P (F > Fa

Mistake SCE (k - 2)(k -1) CMErr

Total SCT k2
-1

*Taken from Gutierrez, H, et all. Analysis of experiments. Chapter Four 132

SCT = SCtrat + SCBl + SCB2 + SCE Equation 75

and the corresponding degrees of freedom are

k2
-1 = (k -1) + (k -1) + (k -1) + (k - 2)(k -1) Equation 76

8.7. Example for Greco-Latin design.

Example of a Greco-Latin square (Pena, 2002)
The performance of three manufacturing processes (A, B, C) is compared under

three experimental conditions (_ α, β,ϑ ) on three different days with three

measurement procedures. The design and results obtained are indicated in the
table.

Experimental Design
Cria43
8
2
 Table 30. ANOVA for the example of a Greco-Latin square design*

Day 1 Day 2 Day 3 Yi…

Method 1 Aα Bβ Cϑ 13
34

Aα Bβ
10 11

Cϑ
Method 2 8 26
8 10

Method 3
Bβ Cϑ Aα 30

and…L 9
27 11
32 31 90

α……………Y..1.=28
β…………….Y..2.=32
ϑ……………..Y..3.=30
Table 30. ANOVA for the example of a Greco-Latin painting*

SOURCES OF SUM OF DEGREES MEDIUM

Fo P-value
VARIABILITY SQUARES OF SQUARES
FREEDOM

Treatments Latin SCtratlatin k-1 CMtratalatin P (F > Fa)

letter Ftralat = CMtratlatin/
CMEr

Lines SCrenglo k-1 CMrenglo P(F > Fa)

Fre = CMrenglol
/CMEr

Columns SCcolumn k-1 CMcolumn

F colmn= P(F > Fa)
CMcolumn/CM
It is
Greek Letter SCtratgiega k-1 CMtaratGreek Ftratlgrieg=CM
Treatments taratgriega P (F > Fa

Mistake SCE (k - 2)(k -1) CMErr

Total SCT k2
-1

Chapter Nine: Factorial Designs

9.1. Factorial designs

In engineering, many experiments are carried out to study the effects produced by
two or more factors.

Experimental Design
Cria43
8
3
Moving these within an experimental domain is costly, and time constraints force
the experimenter to perform only the essential experiments.

In traditional methods, only one factor is varied at a time. This is not optimal
because it involves more experiments than necessary and partial information is
obtained because they do not show the interaction between the factors.
Interactions in an experiment are common and are the most important effects for
understanding the behavior of many systems.

Experimental designs allow several factors to be varied simultaneously but prevent

them from going in the same direction. And they complement each other in such a
way that the information sought is obtained by combining the responses of all of
them.

9.2. Advantages of Factorial Experiments

We can highlight among several the following:

1. Economy in experimental material since information is obtained on several

factors without increasing the size of the experiment.
2. They allow the study of interaction, that is, they determine the degree and
the way in which the effect of a factor is modified by the levels of another
factor.

9.3. Disadvantages of factorial designs

A disadvantage of factorial experiments is that they require a large number of
treatments, especially when there are many factors or many levels of the same
factor. This fact has the following effects:

1. If complete blocks are to be used, it is difficult to find groups of

homogeneous experimental units to apply all the treatments.

2. The cost of the experiment is increased by having many experimental

units; this is minimized by using fractional factorials where only one part of
the entire set of treatments is tested.

Factorial experiments can be run under any type of error control design or
undersampling or with covariates.

In this module, only analyses of two-factor factorial experiments under a DCA or

DBCA will be presented.

Experimental Design
Cria43
8
4
9.4. Factorial arrangement
The set of individual experiments or treatments that are formed when considering
the combination possibilities for the factorial design is shown in the following table;

FACTOR B

FACTOR A A1 B1

Figure 11. A factorial experiment

FACTOR B

FACTOR A

Figure 12. A factorial experiment

9.5. Statistical model for a factorial design

Any observation in the factorial design can be represented by the model;

Yijl = μ+ αi + βj αβ)
+ ( ij + εijk Equation 77

i= 1,……., a number of treatments

j=1,…., b number of blocks
k= 1,2,..n
μ = overall mean, also called β0

αi = effect due to the i-th level of factor A, also called

level of factor B, also called β2 .
β1 β= effect due to the j-th
( αβ) ij = represents the interaction effect in the combination ij, also called
β1β2 .

ε
( )ijk = is the random error that supposes a distribution with zero mean and 2

constant variance σ
The regression model is described by:

Yijl = βo +β1X1 + β2X2 +β1β2X1X2 + εijk Equation 78

Experimental Design
Cria43
8
5
9.6. Hypothesis to be tested

H0: Effect A = 0
Ha: Effect A # 0

H0: Effect B = 0
Ha: Effect B # 0

H0: AB effect = 0
Ha: AB Effect # 0

If effect A= α If effect B= β
Hypotheses can also be stated as follows:

H0 : α1 =α2 = ...αa =0

HA : α#0 for some i

H0 : β1 =β2=...βb= 0
HB : βj#0

H0 :(αβ) ij =0
HAB : (αβ) ij #0 for some ij
Hypotheses that are tested using the analysis of variance (known) technique.

Calculation of the sum of squares of the error.

SCE - SCT = SCA+ SCB+ SCAB Equation 79

Calculation of the sum of squares of the error

SCE=SCT-SCA-SCB-SCAB Equation 80

9.7. ANOVA for factorial (axb) or two-factor design

The analysis of variance is calculated in a similar way to the first paragraphs of
this chapter.

Experimental Design
Cria43
8
6
 9.8. Example of a factorial design
A 3x2 factorial design with 2 replications is run to investigate the shrinkage of a
product after the standardization process.

Table 31. ANOVA for the axb factorial design

Fo p-value
Sum of Degrees of Square s
squares freedom means

EFFECTOA SCA a-1 CMA Fo = CMTRAT / P(F > Fo)

CME

EFFECTB SCB b-1 CMB P(F > Fo)

Fo = CMB / CME

EFFECTAB SCAB (a -l)(b -1) CMTRAT

Mistake SCE ab(n-1) CME

Total SCT abn-1 CMTRAT

The factors investigated are. A: Temperature of the storage warehouse (three

levels, T1, T2 and T3) and content of the preservative (curing material) at two
levels, B1 and B2. The data obtained are shown in the following table:

• Research hypothesis: Temperature and attachments influence the final

size of the product.

• Experiment design: a homogeneous mass for making the product is

divided into six equal portions and the percentage of the selected
ingredient is mixed into each one.

Table 32. NOVA test results for the axb factorial design
Temperature
T1 T2 T3
B1 3,9 4,2 4,3 3.81 3.72 3.84 4 3,8 3,4
Packing B2 3,5 3,8 4,0 3.37 3.50 3,6 3,9 4,1 4,3
** PREPARE THE ANAVA FOR THIS PROCESS AND SHARE IT WITH YOUR CLASSMATES IN THE COURSE
GROUP.

Experimental Design
Cria43
8
7
 Table 33. ANOVA for axb factorial design example

Sum of Fo p-value
Degrees of Ecuac
squares Mean Squares
freedom
Equation
Equation
Ecuac

EFFECTOA

EFFECTB

EFFECTAB

Mistake

Total

* What conclusions would you draw?

9.9. Two-way factorial experiments

In a two-way factorial experiment, two experimental variables are chosen and two
or more levels are selected for each.

For each combination of levels there will be a treatment.

Although factorial experiments are typically longer than univariate experiments,

they typically yield much more information in the time they take.

When analyzing results such as those in Table 9, one must always ask whether
the differences observed are due to the treatments applied or to some other
variable, which may be unknown. The purpose of the experimental design is to
test whether or not there is interaction between the factors.

9.10. Factor in two-way design.

When more than one experimental variable is handled in a single experiment,
these are known as experimental factors.

9.11. Treatment in two-way design.

In a two-way factorial experiment a treatment is a combination of levels of the two
factors. If there are 11 levels of factor 1 and 2 levels of factor 2 then there will be
11, times 11, treatments.

Experimental Design
Cria43
8
8
9.12. Interaction in two-way design.
If the effect of one factor on the response variable is different for different levels of
the other factor, the factors are said to exhibit an interaction. Interactions only
become visible with factorial experiments, which gives them an important
advantage over univariable experiments.

9.13. Steps to reach ANOVA in an experimental design.

To obtain the ANOVA, for example, in table 42, 44 and others to answer specific
questions about an experiment, you must:

1) Get the contrasts. Or comparisons between treatment means.

2)Calculate the effects. See equations 16,17 and 18.
3) The sums of squares. See equations 19,20,21, 22, 23 and 24.
4) Perform the significance test (FO). See ANOVA tables.

Chapter Ten: 2k Factorial Designs

4.1. Coding of variables
Coding consists of assigning numerical or alphanumeric values (codes) to each of
the variables in order to facilitate data processing.

Factor levels for factorial design can be coded with the symbols +1 (top level) and
-1 (bottom level); to provide a uniform and appropriate framework for investigating
the effects of factors (variables) in any experimental region where the actual
values of the factors depend on the variables.
XCodedi = (Xi - X) /Difference Equation 81

Difference = ½ (XM– Xn) Equation 82

(XM– Xn) = difference between the maximum and minimum values.

Xi = is the value of x = the value in the

experiment
X = mean value of = the mean value in experiment k
4.2. FULL FACTORIAL DESIGN 2
For the notation used k = number of factors.

And 2 (base) is the number of levels selected.

In this design only two factors are considered, A and B, each with two levels.

It is customary to refer to the lower level of factors as -1 or also A1.

And the lower level by 1 or also A2.

It is the most appropriate strategy to simultaneously know what effect k factors

have on a response and discover whether they interact with each other.

Interactions are often very common and are sometimes the most important effects,
so knowing them is essential to understanding the behavior of many systems.

This describes the most appropriate experiments to simultaneously determine

what effect k factors have on a response and discover whether they interact with
each other.

Experimental Design
Cria43
8
9
 By not having correlated factors, redundant experiments are avoided.
Furthermore, the experiments complement each other in such a way that the
information sought is obtained by combining the responses from all of them.

Table 34. Algebraic signs for calculating the effects in a design 22

Combination REPETITIONS
PROOF TO B AXB of treatments AVERAGE
AN AN AN And
n
D D D
1 ..
2 AND

1 - - + (1)
2
+ - - to
3 - + - b
4 + + + ab

4.3. Effect of a factor in a design 22

The influence that a factor has on the response variable as the change in
response when a factor changes from its lowest level to its highest level, averaged
over the other factors.

a) The average effect of A is defined as:

A = 1/2n{[ ab-b]+[a-(1)]}

A = 1/2n[ ab+a – b -(1)] Equation 83

b) The average effect of B is defined as:

B = 1/2n{[ ab-a]+[b-(1)]}

B = 1/2n{[ ab + b – a - (1)]} Equation 84

c) The interaction effect of AB is defined as:

AB = 1/2n{[ ab-b]-[a-(1)]}

AB = 1/2n[ ab+(1)-ab] Equation 85

n = number of repetitions or replications

d) The sum of squares of the effects is calculated from the contrasts and is
defined as:

SS=[Effect]2[2k-2]n Equation 86

It can also be calculated

Experimental Design
Cria43
9
0
d1) For factor A this is calculated with:

SCA = {[ab+a -b-(1)]} 2

n22 Equation 87

d2. from B
SCB = {[ab-a +b-(1)]} 2
n22 Equation 88
d3. from AB
SCAB = {[ab+(1)-ab]} 2
n22 Equation 89
d4. Calculation of the sum of total squares

an
SST =∑ ∑ ∑YK2=1-Y2 N Equation 90 i=1 j=1 k=1

N = number of tests

d5. Calculation of the sum of squares of the error

SCE=SCT-SCA-SCB-SCAB Equation 91

4.4. Waste analysis

It is a useful procedure for evaluating the fit of a regression equation. Residuals
are calculated when differences are found between the actual value and the value
of the dependent variable and its predicted value.
In other words, a residual is the part of Y that is not predicted by X.

The mean value of the residuals is zero and since the slope is subtracted in the
process of calculating the residuals, the relationship between these and the X
values is also zero.

Y=β0 +β1X1 +β2X2 + ε Equation 92

4.5. ANOVA for a factorial design
22

With the products of the equations of the previous numeral we must build the
following table.

4.6. Example of a factorial experimental design 22

The steps to follow in solving an engineering exercise or situation where an

experimental design is applied are the following:

1- Problem statement
The performance of a reaction with a new catalyst is to be checked. Therefore, the
aim is to determine to what extent temperature and time influence reaction yield
and how this can be varied to improve it.

4.7. Factors and experimental domain

For this, in accordance with previous experience, the bibliographic review or the
needs of the experimentation (profitability criteria, experimental limitations), it is
necessary to choose which factors are of interest to study and what values they
can take (the experimental domain).

Table 35. ANOVA for a 22 factorial design

Experimental Design
Cria43
9
1
 Sum of Degrees of Mean squares Fo p-value
squares freedom

EFFECTOA SCA 1 CMA P(F > Fo)

Fo = CMA / CME
SCB 1 CMB P(F > Fo)
EFFECTB Fo = CMB / CME
SCAB 1 CMAB Fo = CMAB/ CME P(F > Fo)
EFFECTAB

Mistake SCE 4(n-1) CME

n22-1
Total SCT

Table 35 shows the two factors chosen. Since both factors are continuous, their
experimental domain is expressed by the maximum and minimum values they can
take.

For this purpose, it was agreed that the reaction time should not be less than 8
hours for the process to be profitable, and more than 6 hours to ensure good
performance; in addition, the working temperature range must be higher than
(40ºC) but without exceeding 80ºC.
Table 35 also shows the most common coded notation for continuous factors: the
value –1 is assigned to the lower end of the experimental domain and the value +1
to the upper end 2.
For simplicity, often only – and + are indicated.

It is necessary to define the correspondence between real and coded variables

because the design of experiments describes optimal experimentation using
dimensionless coded variables (X1, X2,...). In this way, mathematical and
statistical tools are general and can be applied to each specific problem.

The most economical experimentation (minimum number of experiments) is one in

which each factor takes only two values (levels). And the one that will provide the
information with the least uncertainty is the one in which these values are the
extremes of the experimental domain, –1 and +1.

This same table 35 shows the matrix of experiments obtained by combining the
two levels of the two factors. Each row is an experiment and each column is a
factor studied.

Table 36. Answers for factorial design example 22

Matrix of Plan of Answer

experiments experimentation (% performance)

x1 x2
Temperature Bullfight 1 Bullfight 2 Bullfight 2
Time (h) (ºC)
1 - - 6 40 49 48 49
2 + - 8 40 54 58 55
3 - + 6 60 70 72 73
4 + + 8 60 78 55 56

Experimental Design
Cria43
9
2
 Table 37. Matrix for factorial design example 22
Combination of Repetitions AND
TO B AXB treatments AVERAGE
PROOF
x1 x2 x1 x2 AN AN AND
D D 3
1 2
1 - - + (1) 49 48 49 48,66666667
2 + - - to 54 58 55 55,66666667
3 - + - b 70 72 73 71,66666667
4 + + + ab 78 55 56 63

Steps to follow

a) The average effect of A is defined as:

A = 1/2n[ ab+a – b -(1)]

A = 1/2n[ ab+a – b -(1)]

A= 7,666

b) The average effect of B is defined as:

B = 1/2n{[ ab + b – a - (1)]}
B = 13

c) The interaction effect of AB is defined as:

AB = 1/2n{[ ab-b]-[a-(1)]}

AB = 1/2n[ ab+(1)-ab] =

AB = -9

n = number of repetitions or replications

Table 38. Analysis of variance for factorial design example 22

Sum of Degrees of Mean Fo p-value

squares freedom squares

176,3333333 1 176,3333333 0,032713984 P(F > Fo)

EFFECTOA

EFFECTB 507 1 507 0,094060433 P(F > Fo)

EFFECTAB 264,4533907 1 264,4533907 0,049062328 P(F > Fo)

Mistake 43121,21328 8 5390,151659

Experimental Design
Cria43
9
3
Total SCT 11

Chapter Eleven: Three-Factor Factorial Design

11.1. Three-factor factorial design with two levels each.
A factorial design involves three factors A, B, C, all with two levels, their
23

combinations are shown in the following table.

Experimental Design
Cria43
9
4
 Table 38. Combination of treatments for a design 23
COMBINATION
OF
RUN TO B AXB AXC BXC AXBXC
TREATMENTS
C
1 - - - + + + - (1)
2 + - - - - + + to
3 - + - - + - + b
4 + + + + - - - ab
5 - - + + - - + c
6 + - + - + - - ac
7 - + + - - + - bc
8 + + + + + + + abc

11.2. Effect of a factor in a design 23

It is the influence that a factor has on the response variable (change in response)
when a factor changes from its lowest level to its highest level, averaged over the
other factors.

Table 39. Algebraic signs for calculating the effects in a design 23

COMBINATION OF TREATMENTS FACTORIAL EFFECT

Yo
AXB
TO B AXB AXC BXC
XC
C
(1) + - - - + + + -
to + + - - - - + +
b + - + - - + - +
ab + + + - + - - -
c + - - + + - - +
ac + + - + - + - -
bc + - + + - - + -
abc + + + + + + + +

a) The average effect of A is defined as: ab ab c ac bc abc

A = 1/2n{[-(1)+a -b+ab-c+ac-bc+abc]} Equation 93

A = 1/2n[ a+ab+abc – b - bc -(1)] Equation 94

b) The average effect of B is defined as:

B = 1/2n{[-(1)-a+b+ab-c-ac+bc+abc]}
Experimental Design
Cria43
9
5
B = 1/2n{[ ab + b+bc+abc – ac-ac - (1)]} Equation 95 c) The interaction effect of
AB is defined as:

AB = 1/2n{[(1)-a-b+ab+c-ac-bc+abc]}
AB = 1/2n{[(1) +abc+ab+c -ab-ac-bc]} Equation 95 n = number of repetitions or
replicates

d) The AC interaction effect is defined as:

AC = 1/2n{[(1)-a+b-ab-c+ac-bc+abc]}
AC = 1/2n{[(1) +b+ac+abc – a -ab-c-bc]} Equation 96 n = number of repetitions or
replicates

e) The ABC interaction effect is defined as:

ABC = 1/2n{[(-1)+a+b-ab+c-ac-bc+abc]}
ABC = 1/2n{[ a +b+c+abc – (1) -ab-ac-bc]} Equation 97 n = number of repetitions
or replicates f) the sum of squares is defined as:

SS= [Effect]2[2k-2]n Equation 98

They can also be calculated.

Calculation of the sum of total squares

an

SST =∑ ∑ ∑YK2=1-Y2 N Equation 98.1

i =1 j=1 k=1

g) Calculation of the sum of squares of the error.

SSE= SST-SSA-SSB-SSAB Equation 99 h) Residue analysis

Y=β0 +β1X1 +β2X2 + ε Equation 100

Experimental Design
Cria43
9
6
11.3. Analysis of variance for a factorial design 23

23
Table 40. Calculating the Analysis of Variance for a Factorial Design

Fo p-value
Sum of Degrees of Square s
squares freedom means

EFFECTOA SCA 1 CMA P(F > Fo)

Fo = CMA / CME

EFFECTB SCB 1 CMB P(F > Fo)

Fo = CMB / CME

EFFECTC SCC 1 CMC P(F > Fo)

Fo = CMC / CME
Fo = CMAB/ CME
EFFECTAB SCAB 1 CMAB P(F > Fo)

Fo = CMAC / CME
AC EFFECT SCAC 1 CMAC P(F > Fo)

Fo = CMBC / CME
BC EFFECT SCBC 1 CMBC P(F > Fo)

Fo = CMABC /
ABC EFFECT SCABC 1 CMABC CME P(F > Fo)

Mistake SCE (n-1)

23
CME

Total SCT n23-1

11.4. Coefficients of determination

These coefficients measure the proportion or coefficient of variability in the
experimental data that is explained by the model considered.

R2 = { [ SCTotal -SCError /SCTotal]100} Equation 101

R2
= [SCModel/SCTotal]100} Equation 102

11.5. Example of a factorial design

23

A company producing chocolate drinks has had problems with the viscosity of a
certain chocolate drink. It is believed that three ingredients

Experimental Design
Cria43
9
7
 added in small quantities can solve this problem, so it is necessary to explore the
situation; for this purpose, an experiment 23 with two replicas is run. With the
results obtained:

Table 41. Viscosity data of the developed beverage.

INGREDIENT A INGREDIENT B INGREDIENT C GOO
y1 y2 AND. m
-1 -1 -1 13,3 13,7 13,4985184
+1 -1 -1 14,7 14,4 14,5492268
-1 +1 -1 14,6 14,5 14,5499141
+1 +1 -1 14,3 14,1 14,1996479
-1 -1 +1 16,9 17,2 17,0493402
+1 -1 +1 15,5 15,4 15,4499191
-1 +1 +1 17,0 17,1 17,0499267
+1 +1 +1 18,9 19,0 18,949934

Table 42. Average and factors calculated for viscosity example

Average 15,6621 +/- 0,61255
Factor A 0,250257 +/- 1,2251
Factor B 1,0506 +/- 1,2251
Factor C 2,92545 +/- 1,2251
AB factor 0,524613 +/- 1,2251
AC Factor -0,099964 +/- 1,2251
BC Factor 0,699696 +/- 1,2251

Table 43. Analysis of variance for developed beverage viscosity.

Sum of Degrees of Mean squares F p
squares freedom
Factor_A 0,125257 1 0,125257 0,04 0,8717
Factor_B 2,20754 1 2,20754 0,74 0,5487
Factor_C 17,1166 1 17,1166 5,70 0,2525
AB 0,550439 1 0,550439 0,18 0,7424
AC 0,0199856 1 0,0199856 0,01 0,9482
BC 0,979149 1 0,979149 0,33 0,6696
Total error 3,00174 1 3,00174
Total (corr.) 24,0007 1

Experimental Design
Cria43
9
8
 Figure 13. Steps to calculate the ANOVA in a 2k design

For this case, there are zero effects that have a p-value less than 0.05, which
tells us that they are significantly different from zero at the 95% confidence level.

Table 44. Regression coefficients of the beverage viscosity data equation

developed in the previous example.

Constant = 15,6621 Factor_A = 0.125129

Factor_B = 0,525302 Factor_C = 1.46273
AB = 0,262307 AC = -0,049982
BC = 0,349848 R2
= 87,4931
The fitted equation for the model is
VISCOSITY = 15.6621 + 0.125129*Factor_A + 0.525302*Factor_B +
1.46273*Factor_C + 0.262307*Factor_A*Factor_B -

The values of the variables being specified in their original form.

Experimental Design
Cria43
9
9
 0.049982*Factor_A*Factor_C + 0.349848*Factor_B*Factor_C.
Equation 37.
R-squared (adjusted for degrees of freedom) = 12.4516 percent
Standard error = 1.73255

YO[
Absolute mean error = 0.61255

Table 45. Values for r2 , the standard error for the example of the viscosity of the
developed beverage.
The statistical value of R2 indicates that the adjusted model explains 87.4931%
of the viscosity variability, it is more appropriate to compare models with different
numbers of independent variables is 12.455%

The estimate of the standard error shows that the standard deviation of the error
of the residuals is 1.73255.

The mean absolute error (MAE) was 0.61255 which is the average value of the
residuals.

Table 46. Estimated results for the viscosity data with the regression equation
found for the previous example.

Data
Observed Lower 95.0%
Fixed value CL for Mean Upper 95.0% CL
value
1 13,4985 14,1111 -6,48132 34,7035
2 14,5492 13,9367 -6,65571 34,5291
3 14,5499 13,9374 -6,65502 34,5298
4 14,1996 14,8122 -5,78019 35,4046
5 17,0493 16,4368 -4,1556 37,0292
6 15,4499 16,0625 -4,52992 36,6549
7 17,0499 17,6625 -2,92991 38,2549
8 18,9499 18,3374 -2,255 38,9298

11.6. Presentation of results.

We suggest interpreting the results in the following order and with the following
recommendations:

Experimental Design
Cria43
1
0
 1) Average value
2) Main effects
3) Interaction effects of two factors
4) Do not forget that interaction exists when the effect of a factor is different at
different levels of another factor(s). This can be easily understood if we evaluate
the effect of each factor in pairs of experiments.
5) Three-factor interaction effect

For example, for the viscosity exercise, the following should be presented:

Figure 16. Mean effects for viscosity example.

Figure 17. Interactions for viscosity example.

11.7. Response surface methodology. Various authors.
The experimental strategy or experiment planned to model and determine how
the levels of one factor respond when compared to those of another is known as
a response surface.

For the example of the viscosity exercise the response surface obtained is the
following

Figure 18. Interactions for viscosity example.

11.8. Experimental region

It is the space delimited by the ranges of experimentation used with each

Experimental Design
Cria43
1
0
 factor.

11.9. Region of operability

Set of points where the equipment or process can be operated.

11.10. Orthogonal polynomial.

They are regression equations in which each one is associated with an
exponent of the independent variable, for example X, X1, X2 and all are
mutually independent, that is, they are orthogonal.

Chapter Twelve: Regression Analysis

12.1. Regression analysis.
Regression analysis is useful for finding out the likely form of the
relationship between variables, and when using this method, the ultimate
goal is usually to predict or estimate the value of one variable that
corresponds to a given value of another.

Based on the results obtained from the analyses carried out on the
representative sample, extracted from a given population, the researcher
will be able to understand and thoroughly know the nature and behavior
of the population in which he is interested.

12.2. Simple linear regression. (2, 3, 7 and various authors)

Regression analysis is useful for finding out the likely form of the
relationship between variables, and when using this method, the ultimate
goal is usually to predict or estimate the value of one variable that
corresponds to a given value of another.

On the other hand, correlation analysis refers to the measurement of the

intensity existing between the relationships of the variables. When
calculating the correlation coefficients of a set of data, the focus is on the
degree of compatibility between the variables that make up the problem.

Table 48. Equations for estimating a simple linear regression

Y = a+bX Equation 11

b={∑XY-[∑ X
Equation 12
∑ Y n]} ∑X2-[∑X2 n]

a=∑Yb ∑ X n Equation 13

r = {∑ XY -[ ∑ X∑Y n]}

//[) ∑ X -[∑ n ][∑Y-

X2
∑ Y2
n

Experimental Design
Cria43
1
0
Equation 14

It is also represented as

Y = β0 + β1 X + ε
Equation 15
Where
β0 = a = inter section β1 = b = slope

ε = change in error. Random variable

12.3. Correlation coefficients.

A correlation is the degree of relationship between two or more variables
that seeks to determine to what extent a line or other equation describes
or explains that relationship.

They are within a range that goes from -1 to 1 range that explains the
scope and assessment of said coefficient and the direction of the
correlation. Table 49.

Table 49. Degree of association of the correlation coefficients for a set of

data. (various authors)

CORRELATION COEFFICIENT INTERPRETATION

-1.00 Perfect negative correlation
- 0.95 Strong negative correlation
- 0.50 Moderate negative correlation
- 0.10 Weak negative correlation
0.0 There is no correlation
+ 0.10 Weak positive correlation
+ 0.50 Moderate positive correlation
+ 0.95 Strong positive correlation
+ 1.00 Perfect positive correlation
The degree of association, the closer it is to unity, in one direction or
another, the greater the correlation.

Generally, in most research works, the aim is to establish a linear

correlation, whether positive or negative; it cannot be considered that all
relationships between two variables (X and Y) form a straight line. (see
next chapter).

There are many correlations of a curvilinear nature that indicate that one
variable increases as the other increases until the relationship itself is
reversed, such that one variable finally decreases while the other
continues to increase, that is, the relationship between X and Y that is
initially positive becomes negative, or that which is initially negative
becomes positive; for example in a chemical reaction or in a biochemical
reaction.

12.4. Other simple linear regression models. (2, 3, 7 and various

authors)

Experimental Design
Cria43
1
0
 Table 50. Simple linear regression models
EQUATION MODEL
Y = ea + bX Exponential Equation 16
Y =1 (a+bX)
The reciprocal Equation 17
Y = aXb) Multiplicative Equation 18
Y = a + b ln(X ) Logarithmic Equation 19
Y=a+b (X) Reciprocal
Equation 20

12.5. Multiple linear regression.

Multiple linear regression analysis consists of generating regression
models with more than one independent variable (X1, X2, X3…..); the
multiple linear regression model is represented by:

Y = β + β X + β X +.............+βX +ε Equation 30
0 11 22 kk

k = number of variables (Xs)

12.6. Multiple correlation.

If more than one independent variable is needed in the previous equation
(X1, X2...) the term "multiple correlation" is applied.

Calculations for this type of correlation are more expensive and the
interpretation of its results more complex. The inclusion of additional
variables increases the data needed and may also add considerable cost
to the study.

The main reasons for changing the analysis from one to two or more
independent variables are:

There is a logical relationship, and a single independent variable does

not provide a sufficiently satisfactory correlation coefficient.

The multiple correlation coefficient represents the proportion of variation

in Y that is explained by the selected set of explanatory variables. (3, 7, 8,
27 and various authors)

12.7. Data transformation. (3, 7, 8, 27 and various authors)

In practice, many of the experimental processes do not follow a normal
distribution with constant variance and although the ANOVA procedure is
robust and allows for moderate deviations, performing a statistical
analysis with results that do not meet the assumptions about the
statistical model can lead to an erroneous conclusion and decision.

For this purpose, through an appropriate transformation, a variable that

is not normally distributed can be made to have an almost normal
distribution; for example, a change in scale can vary the mean and
variance of the variable as well as its relationship with respect to other
variables, just as the shape of the distribution of a variable changes with
the scale.

Experimental Design
Cria43
1
0
 Populations with unequal variances can be converted to homogeneous
variances by an appropriate transformation.

The most commonly used transformations are presented in the following

table:

Table 51. Transformations used for data

Appropriate transformation Type of transformation
Y'
= ln(Y) = log10(Y)
OY'
Logarithmic transformation

Y' = Y -1/2
Square root transformation
Y' =Y-1 Reciprocal transformation

12.8. Multiple correlation.

If more than one independent variable is needed in the previous equation
(X1, X2...) the term "multiple correlation" is applied.

Calculations for this type of correlation are more expensive and the
interpretation of its results more complex. The inclusion of additional
variables increases the data needed and may also add considerable cost
to the study.

The main reasons for changing the analysis from one to two or more
independent variables are:

There is a logical relationship, and a single independent variable does

not provide a sufficiently satisfactory correlation coefficient.

The multiple correlation coefficient represents the proportion of variation

in Y that is explained by the selected set of explanatory variables. (3, 7, 8,
27 and various authors)

12.9. Analysis of Covariance.

Analysis of covariance is a very important procedure in experimentation,
but unfortunately it is not used frequently.

It uses analysis of variance and regression to eliminate the variability that

exists in the independent variable X; it also adjusts treatment means and
thus estimates much better the effect of the independent variable X on
the dependent variable Y.

The independent variable X is an observation made on each

experimental unit before applying the treatments, and indicates to a
certain degree the final response of the experimental unit.

12.10. The objectives of analysis of covariance are:

a) Decrease the experimental error with the respective increase in
precision of the experiment.
b) Adjust treatment averages.
c) Better interpret the results of the experiment, especially as they
relate to the nature of the treatment effects.
d) Estimate the value of units lost in experiments.

Experimental Design
Cria43
1
0
 12.11. Statistical software for problem solving.
Among the programs suitable for data management, processing and
analysis we mention the following:

Excel, Minitab, Statgraphics, SAS, Curve Expert; etc. However, with the
use of a calculator it is also possible to carry out the vast majority of
operations involved in data processing.

Chapter Thirteen: Non-Experimental Designs

13.1. Non-experimental designs. Various authors
It is a design that is based on direct observation, interviews and the
review of documentary sources; without exercising control or
manipulation of the variables under study, but rather observing the
development of the situations and, based on a careful analysis,
attempting to extract explanations of certain validity. Since there is no
control of the variables, many sources of error may arise that cast doubt
on their validity.

13.2. Recommendations
Scientific observation must follow some basic principles:

1. It must have a specific purpose.

2. It must be planned carefully and systematically.
3. A careful record of it must be kept in writing.
4. Its duration and frequency must be specified.
5. It must follow the basic principles of validity and reliability.

13.3. Proposed exercises

Solve the following exercises and share them with your tutor and
classmates during face-to-face classes.

Exercises to develop

And socialize at the face-to-face times agreed upon with the group.

1- An engineer presents the following factorial design. Answer the

following sections without using computer software, that is, do the
operations requested manually.

a) What is the name of this design and why?

b) What treatments does this design have, how many replicas?
c) In total, there were twelve experimental runs that were carried out.
Indicate in what order they should have been run and explain why.
d) Point out the effects that can be studied through this design.

REPLICA
TO B Yo ii iii Total
- - 66 74 67 (1) = ?
+ - 61 76 79 (a) = ?
- + 71 70 83 (b) = ?
+ + 75 67 80 (ab) = ?

Experimental Design
Cria43
1
0
 e) Obtain the contrasts for the main effects of A and B and for the
interaction.
f) Make the graph of the interaction between factors A and B and
interpret it in detail.
In fact, factor B has an influence on factor A.

2- A food engineer is asked to analyze the breaking strength (Y) of a

food package, for which he uses a numerical scale.
Examines three factors, each at two levels, X1 = plastic material, X2 =
humidity, X3 = direction of scratch.
Decide to obtain three observations or replications in each combination,
the same ones shown in the following table:

a) Perform the ANAVA for this data.

b) Interpret the significant data and what conclusion would you draw as
an engineer?

X1 X2 X3 Breaking strength
- - - 3,25 3 2,3
+ - - 2,1 2,9 4,1
- + - 4,5 3,9 3,8
+ + - 5,1 5,1 3,2
- - + 3,7 5,7 4,9
+ - + 3,2 5,3 4,3
- + + 5 2,8 4
+ + + 4,7 2,7 3,5

LITERATURE

1. Bavaresco, Aura (1979): Research techniques. Edit. Scott,

Foresman & Co. 4th edition. Gienview, Illinois, USA. (1st
edition: 1974).
2. Berenson, Mark; Levine, David; Krehbiel, Timothy. Statistics
for administration. Edit, pearson, 2nd Edition, mexico 2001,
734p
3. BickinG A. C. Some Uses of statistics in the Planning of
Experiments. Industrial Quality Control, Vol. 10 No. 4, January
1954.
4. Bunge, Mario (1980): Science and development. Edit.
Twentieth Century. Buenos Aires, Argentina.
5. Christensen, Howard. Statistics step by step. Edit Trillas. 5th
edition, 1990. Mexico. 682 p.
6. Cochran, William; Cox, Gertrude. Experimental designs. Edit,
Trillas, 5th reprint. Mexico 2001, 661p.
7. Costamagna, Alicia, M. Conceptual maps as an expression of
interrelation processes to assess the evolution of knowledge in
university students. Science Teaching, 2001.19, (2), 309-318.
8. COX DR Planning of Experiments. John Wiley and Sons, Inc.
New York, 1978. 700p.
9. Cortes, Manuel; Sanchez, William. Technological research and
development project. National Open and Distance University.

Experimental Design
Cria43
1
0
 Faculty of Basic Sciences and Engineering; Bogota; 2001;
313.Pages.
10. Dawson, Beth; Trapp, Robert. Medical biostatistics. Edit,
modern manual. 3rd Edition. Mexico 2002, 435p.
11. FAOSTAT.
12. FSTA. Food science technology abstracts
13. Hernandez, Sampieri, Roberto. Research methodology. Edit
Mac Graw hill. Mexico, 2003, 692 p.
14. Hinkelmann, Klaus. Design and analysis of experiments. Vol I,
edit. john wiley & sons, inc.1994. 495p.
15. Howard, B, Cristensen. Statistics step by step. Edit, Trillas, 3rd
edition. Mexico 2004, 682p.

16. COLOMBIAN INSTITUTE OF TECHNICAL STANDARDS

AND CERTIFICATION Compendium of Colombian Technical
Standards on Documentation, Theses and other degree works.
Santa Fe de Bogotá: ICONTEC, 1996.
17. Kempthorne O. The Design and Analysis of Experiments. John
Wiley and Sons. New York, 1952, p.10.
18. Kuehl, O, Robert. Design of experiments. Thomson learning,
2nd edition. Mexico.
19. Mendez I. General Guidelines for Planning Experiments.
Monograph No. 15, Vol. 15 IIMAS. 1980.
20. Montgomery, Douglas. Design and analysis of experiments.
Ibero-American publishing group.1991. 589p.
21. Novak, Joseph, D; Gowin, D, Bob. Learning to learn. Edit.
Martinez Roca, DL 1988. - 228 p.
22. Padron, Corral, Emilio. Experimental designs with application
to agriculture. Trillas Publishing House, 1996.
23. Pereira, Manrique. Collection of readings on research theory
and practice. Volume I and II. UNAD, 2001.
24. Ostle B. Applied Statistics. Limusa-Wiley, Mexico, 1975,
Chap. 10
25. Montgomery, Douglas. Design and analysis of experiments.
Ibero-American Publishing Group, 1st Edition, Mexico 1991,
589p.
26. Sabino, Carlos A. The research process. Ed. Lumen -
Humanitas. Argentina (1996).
27. Tamayo and Tamayo, Mario. The process of scientific
research. 3rd ED.. ED. Limusa SA Mexico (1998)
28. Valdes de Leon, Gustavo. On the possibility of
experimentation in design. May 2004.

Web page addresses.

www.virtual.unal.edu.co/cursos/ciencias/2000352/ -

www.fao.org/docrep/T0848S/t0848s03.htm.pci204.cindoc.csic.es/
tesauros/Tes_ Psic/HTML/PSI_D6.HTM

http://www.sas.com/software/index.html

Experimental Design
Cria43
1
0
 www.educachile.cl/eduteca/estadistica/esc_medic.htm

en.wikipedia.org/wiki

Experimental Design
Cria43
1
0
 APPENDICES
Table 1.

Table E.1 Random number table

Column
OOOO 00001 11111 11112 22222 22223 33333 3333
Row 67890 12345 67890 12345 67890 12345
O 4
0234
01 49280 88924 35779 00283 81163 07275 89863 8
02 61870 41657 07468 08612 98083 97349 20775 4509
03 43898 65923 25078 86129 78496 97653 91550 0807
04 62993 93912 30454 84598 56095 20664 12872 S464
05 33850 58555 51438 85507 71865 79488 76783 3170
8 9247
06 97340 03364 88472 04334 63919 36394 11095
07 70543 29776 10087 10072 55980 64688 68239 020461
08 89382 93809 00796 95945 34101 81277 66090 8887
09 37818 72142 67140 50785 22380 16703 53362 4494
10 60430 22834 14130 96593 23298 56203 92671 1592
5 1367
11 82975 66158 84731 19436 55790 69229 28661
12 39087 71938 40355 54324 08401 26299 49420 5 5920
13 55700 24586 93247 32596 11865 63397 44251 4318
14 14756 23997 78643 75912 83832 32768 18928 5707
15 32166 53251 70654 92827 63491 04233 33825 6966
2 0487
16 23236 73751 31888 81718 06546 83246 47651
17 45794 26926 15130 82455 78305 55058 52551 7 4718
18 09893 20505 14225 68514 46427 56788 96297 7882
19 54382 74598 91499 14523 68479 27686 46162 8355
20 94750 89923 37089 20048 80336 94598 26940 3685
8 8294
21 70297 34135 53140 33340 42050 82341 44104
22 85157 47954 32979 26575 57600 40881 12250 9 7374
23 11100 02340 12860 74697 96644 89439 28707 2581
24 36871 50775 30592 57143 17381 68856 25853 3504
25 23913 48357 63308 16090 51690 54607 72407 5553
8 5705
26 79348 36085 27973 65157 07456 22255 25626
27 92074 54641 53673 54421 18130 60103 69593 4 4946
28 06873 21440 75593 41373 49502 17972 82578 1636
29 12478 37622 99659 31065 83613 69889 58869 2957
30 57175 55564 65411 42547 70457 03426 72937 8379
2 7300
31 91616 11075 80103 07831 59309 13276 26710
32 78025 73539 14621 39044 47450 03197 12787 0 4770
33 27587 67228 80145 10175 12822 86687 65530 4932
34 16690 20427 04251 64477 73709 73945 92396 6826
35 70183 58065 65489 31833 82093 16747 10386 5929
3 6725
36 90730 35385 15679 99742 50866 78028 75573
37 10934 93242 13431 24590 02770 48582 00906 7 5859
38 82462 30166 79613 47416 13389 80268 05085 9666
39 27463 10433 07606 16285 93699 60912 94532 9563
40 02979 52997 09079 92709 90110 47506 53693 4989
2 6732
41 46888 69929 75233 52507 32097 37594 10067
42 53638 83161 08289 12639 08141 12640 28437 7 0926
43 82433 61427 17239 89160 19666 08814 37841 1284
44 35766 31672 50082 22795 66948 65581 84393 1589
45 10853 42581 08792 13257 61973 24450 52351 1660
2 4843
46 20341 27398 72906 63955 17276 10646 74692
47 54458 90542 77563 51839 52901 53355 83281 8 1917
48 26337 66530 16687 35179 46560 00123 44546 7989
49 34314 23729 85264 05575 96855 23820 11091 79821
50 28603 10708 68933 34189 92166 15181 66628 5859
continues 9

Taken from Gutierrez.

Experimental Design
Cria43
1
1
 Table 2. Critical values of F

cotttítination of degrees
For M «a given -—
of freedom in mimeradar and erzonzürzador, jr *,
The element represents the critical value of / 3— 07.05
higher specified (c) n ‘ -----
corresponding to a tail urea
F TO - Fuc, gl,gu}
2=----

Denominato Nurerador, gli

r
g,1 1 2 3 4 5 6 7 8 9 re t2 15 20 24 30 40 60 120 to

1 161.4 199.5 215.7 224.6 230.2 234,0 236.8 238.9 240.5 241.9 243.9 245.9 248.0 249.1 250.1 2511 2521 253.3 254.3
2 18.51 19.00 19.16 19.25 19.30 15.33 19.35 19.37 18,38 19.40 19.41 19.43 19,45 19-45 19.46 19.47 19.48 19.49 19.50
3 10.13 8.55 6-28 9,12 0.01 8.94 8.E9 8.05 B.01 8,79 8.7d 8.70 #,66 B.B4 B.5 8.69 «.57 6.55 8.53
7.71 6.94 6.59 6.39 616 6,16 6.09 6.04 6.00 5.96 6.9T 5.86 5.90 5.77 5.75 5.72 5.59 5,66 5.63
5 6.51 579 5.41 6.19 5.05 4.85 4.B8 4,82 4,77 4.74 4.68 4.52 4.56 4.53 450 446 4.43 4.40 4.38
6 5.99 514 4.76 4.53 4.39 4.28 4.21 4.15 4.t0 4.06 4.00 3.94 3.07 3.84 3.81 3,77 3.74 3.70 3.B7
7 5-59 4.74 435 4.12 3.97 3.87 3.79 3.73 3.68 3.64 3,57 3,61 3.44 3.41 3.3B 314 3.3a 3.27 3.23
B 5,32 4.46 4.07 3.84 3.69 3.66 3.50 3.44 3.39 3.35 3.28 3.22 3.15 3,12 3.06 3.04 3,01 2.07 2.53
9 5.12 4.26 3.86 3.63 3.49 3,37 3.28 3,23 3.18 3-14 3.07 3.01 2.94 2.90 2,88 2.83 2.7» 2.75 2.71
10 4.96 4 10 3-71 348 3.33 3.22 3.14 3.07 3.02 2.98 291 2,85 2,77 2.74 2.70 2.88 2.62 2,58 2.64
4.84 3.88 3.58 336 3.20 3.09 3.01 2.85 2.90 2.85 2.79 2.72 2.65 2.61 2.57 2.53 2,49 2,45 2.40
11 4.75 3.89 3.49 3.26 3.11 3.00 2.91 2.85 2.80 2.76 2.69 2.62 2.54 251 2.47 2.43 2.38 2.34 2.30
13 4.67 3.81 3.41 3.18 3.03 2.92 2.83 2,77 2,71 2.6? 2.60 2.53 2.46 2,42 2.38 2.34 2,30 2,25 2.21
14 4.60 3.74 3.34 3.11 2.96 2.85 2.76 2,70 2.85 2.60 2.63 2.46 2-39 2.35 2.31 217 212 2 18 2.13

15 4,54 3.68 329 3,06 790 2,79 271 2.64 2.69 2.51 2.48 2.40 2.23 2.29 215 2.20 2.15 2.11 2 07
16 4,49 3.63 3.24 3,01 285 2.74 2.86 2.59 2.54 2.49 2.42 2.35 2,28 2.24 2.19 2.15 2.11 2.08 2.01
4.45 3.59 3.20 2.96 2.81 2.70 2.61 2.55 2.49 2.46 238 2.31 2.23 2.19 2.15 2.10 2.05 2.01 1 96
16 4,41 3.55 3.16 2.93 2,77 2.66 2.5S 2.51 2.46 2.41 2.34 217 2.19 2.15 2:11 2.06 2.02 1 97 1.92
18 4,38 3.57 3,13 2.90 2.74 7.63 2,54 24a 2.42 2.38 2.31 2.23 2 18 211 2.07 2.03 1,98 1.93 1.88
20 4.35 3.49 3.10 287 271 2.60 2.51 2.45 2.39 2.35 2,28 210 2.12 2.08 2,04 1.99 THE 1,90 1.84
2li 4.32 3.47 3.07 2.8-4 2.66 2.57 2.49 2.42 2.37 2.32 2.25 2,18 2.10 2,05 2.01 196 192 1.87 1.81
22 4.30 3,44 3.05 2.82 266 2.S5 2.46 2,40 2,34 2.30 2.23 2.15 2.07 2.03 1- 1.94 1.89 1,84 1 78
4.28 3.42 3.08 2.B0 2.54 2.53 2.44 2,37 2,32 217 2.20 2.13 2.05 2.01 '.98 1.91 1.86 1.81 i 76
24 4.26 3.40 3.DI 2.78 2.62 2.51 2.42 2.35 2.30 2.2 2.18 2.11 2,03 1.98 194 189 1.84 1.73 173
25 4.24 3.39 2.99 2.76 2,80 2,49 2,40 2-34 2-20 5
214 2.16 2.09 2.01 1.96 1.92 1.87 1.82 1.77 171
473 a.37 2.96 2.74 2.59 2.47 2.39 2-32 2.27 212 2.15 2.07 1.39 1,35 1.80 1.85 1.BD 1.75 1,69
4.21 3.35 2.96 2.73 2.57 2.46 2.37 2.31 2.25 210 2.13 2.08 1.97 1.93 1.88 1.84 1.79 1.73 1.51
4 34 2.95 2.71 2.56 2.45 2.38 2.29 2.24 2.19 2.12 2.04 1,96 1.91 187 1.82 1.77 171 1.85
2.93 2 70 2.55 2.43 2.36 2.28 2.22 2.18 2.10 2.03 1.94 1.90 1.65 1.81 1.75 1.70 1.84
30 4.17 3.32 2.82 2.69 2.53 2.42 2.33 2.27 2.21 2.1 2.09 2.01 1,93 I.BS RE 1.79 1,74 1.88 1 82
4.DB 3-23 2 E4 3.61 2.45 2.34 2.25 2.18 2.12 6
2.08 2.00 1,92 1,84 1.79 AD
1.74 1.69 1.64 1 58 1 51
IZO 4.00 3.15 2.76 2,53 237 7,25 2.17 2.10 2.04 199 1.92 1,84 1,75 1.70 1.65 1.59 1.53 1.47 139
3.92 3.07 2.88 2.46 2.29 2.17 2,09 2,02 1,96 1.9 1.63 1.75 1.66 161 155 1.50 1,43 L35 116
3.81 3.00 2.60 2.37 2.21 2-10 2-01 1.54 1.88 1 1,75 W 1.57 1.62 1.48 1.30 1.35 1.22 1.00
1,8

Experimental Design
Cria43
1
1
 2

Table 3. Critical values of 1

For a given number of degrees of freedom, the element represents the critical value of x- that
corresponds to a specified upper limit
area («)

Upper tail areas

freedom .995 99 .975 95 .90 ,75 .25 10 .05 .025 .01 005
1 0.001 0.04 0.016 0.102 1.323 2.706 3 841 5.024 6.635 7.879
2 0,010 0.020 0.051 0.103 0.211 0,575 2 773 4.605 5 991 7.378 9.210 10.597
3 0.072 0.115 0.216 0 352 0.584 1,213 4 108 6.251 7 815 9.348 11 345 12.838
4 0.207 0.297 0.484 0.711 ’ 064 1.923 5 385 7 779 9488 11 143 13277 14 860
5 0,412 0.5541 0.831 1 145 1 610 2.675 6 626 9236 11.071 12.833 15 086 16.750
6 0676 0872 1.237 1.635 2.204 3.455 7 841 10 845 12.592 14.449 16.812 18.548
7 0 989 1.239 1.690 2,167 2.833 4.255 9.037 12.017 14 067 16.013 18.475 20.278
3 1.344 1.646 2.180 2.733 3.490 5.071 10.219 13.362 15.507 17.635 20.090 21.955
9 1 735 2088 2,700 3,325 4,168 5,899 11 389 14 684 16.919 19.023 21.666 23.589
10 2 156 2 558 3,247 3.940 4865 6.737 12 549 15 987 18307 .20,483 23.209 25. isa
go 2.603 3.053 3.816 4,575 5 578 7.584 13.701 17.275 19675 21.920 24.725 26.757
12 3.074 3.571 4.404 5.226 6.304 8.438 14 845 18.549 21 026 23.337 26.217 26.299
13 3.565 4.107 5.009 5.892 7.042 9.299 15.984 1 9.812 22.362 24.736 27.688 29.819
14 4.075 4.660 5.629 6.571 7.790 10.165 17.117 21.064 23.585 26 119 29.141 31,319
35 4,601 5.229 6.262 7.281 8.547 11.037 18.245 22307 24.996 27.488 30.578 32.801
16 5.142 5.812 6908 7.962 9.312 11.912 19.369 23.542 26.296 20,845 32.000 34.267
17 5.697 6 408 7.564 8.672 10.085 12 792 20.489 24.769 27.587 30.191 33.409 35.718
18 6.265 7.015 8231 9390 10.865 13.675 21,605 25.989 28.869 31.526 34.805 37.156
19 6 844 7.633 8.907 10.117 11.651 14.562 22.718 27.204 30.144 32.852 36.191 38.582
20 7.434 B.260 9.591 10.851 12.443 15.452 23.823 28412 31.410 34 170 37.566 39997
21 8.034 8.897 10.283 11.591 13.240 16.344 24.935 29.815 32.671 36 479 38 932 41.401
22 8.6413 9.542 10.382 12.338 14,042 17.240 26039 30813 33.924 36 781 40.289 42.796
23 9260 10.196 11 689 13 091 14.848 18.137 27.141 32.007 35 172 38 076 41.638 44 181
24 9 886 10,856 12.401 13.848 15.659 1 9.037 28.241 33 196 36,415 39364 42.980 45,559
25 10,520 11.524 13,120 14.511 16,473 19 939 29339 34.38? 37 652 40.646 44,314 46,928
26 11.160 12,198 13.844 16.379 117.292 20.843 30.435 36,563 38.885 41 923 45.642 48.29©
27 11.808 12.879 14.573 16.151 18.114 21 749 31 528 38 741 40.113 43,194 46 963 ¿9.646
28 12.461 13.565 15.308 16.928 18.939 22.657 32.820 37,918 41,337 44.451 48.278 50,993
29 13 121 14.257 16.047 17 708 19.768 23 567 33 711 39 087 42 557 45.722 49.588 52 336
30 13.787 14.954 10 791 18 433 20 599 24.478 34.800 40.256 43 773 46 979 50.892 53.672

For rafores my gratides tic ¡all fe Hhenod te i>ttfdf roast the expression Z = V 2x2 - V2(gt) - 1 y <1 tíríñ of
the tail siqwrinr that n-'nttn can be abum-r the table of ia standard normal disfñbiicirfn fTiíhlu
eu E2n),

Experimental Design
Cria43
1
1
 Table 4. Duncan test 5%
r
^f)
p

F 2 3 4 5 6 7 3 9 10 20 50 100

1 18.00 18.00 18.00 18.00 18.00 18.00 18.00 18.00 18.00 18.00 18.00 18.00
2 6.09 6.09 6.09 6.09 6.09 6.09 6.09 6.09 6.09 6.09 6.09 6.09
3 4.50 4.50 4.50 4.50 4.50 4.50 4.50 4.50 4.50 4.50 4.50 4.50
4 3.93 4.01 4.02 4.02 4.02 4.02 4.02 4.02 4.02 4.02 4.02 4.02
5 3.64 3.74 3.79 3.83 3.83 3.83 3.83 3.83 3.83 3.83 3.83 3.83

6 3.46 3.58 3.64 3.68 3.68 3.68 3.68 3.68 3.68 3.68 3.68 3.68
7 3.35 3.47 3.54 3.58 3.60 3.61 3.61 3.61 3.61 3.61 3.61 3.61
8 3.26 3.39 3.47 3.52 3.55 3.56 3.56 3.56 3.56 3.56 3.56 3.56
9 3.20 3.34 3.41 3.47 3.25 3.52 3.52 3.52 3.52 3.52 3.52 3.52
10 3.15 3.30 3.37 3.43 3.46 3.47 3.47 3.47 3.47 3.48 3.48 3.48

11 3.11 3.27 3.35 3.39 3.43 3.44 3.45 3.46 3.46 3.48 3.48 3.48
12 3.08 3.23 3.33 3.36 3.40 3.42 3.44 3.44 3.46 3.48 3.48 3.48
13 3.06 3.21 3.30 3.35 3.38 3.41 3.42 3.44 3.45 3.47 3.47 3.47
14 3.03 3.18 3.27 3.33 3.37 3.39 3.41 3.42 3.44 3.47 3.47 3.47
15 3.01 3.16 3.25 3.31 3.36 3.38 3.40 3.42 3.43 3.47 3.47 3.47

16 3.00 3.15 3.23 3.30 3.34 3.37 3.39 3.41 3.43 3.47 3.47 3.47
17 2.98 3.13 3.22 3.28 3.33 3.36 3.38 3.40 3.42 3.47 3.47 3.47
18 2.97 3.12 3.21 3.27 3.32 3.35 3.37 3.39 3.41 3.47 3.47 3.47
19 2.96 3.11 3.19 3.26 3.31 3.35 3.37 3.39 3.41 3.47 3.47 3.47
20 2.95 3.10 3.18 3.25 3.30 3.34 3.36 3.38 3.40 3.47 3.47 3.47

30 2.89 3.04 3.12 3.20 3.25 3.29 3.32 3.35 3.37 3.47 3.47 3.47
40 2.86 3.01 3.10 3.17 3.22 3.27 3.30 3.33 3.35 3.47 3.47 3.47
60 2.83 2.98 3.08 3.14 3.20 3.24 3.28 3.31 3.33 3.47 3.48 3.48
100 2.80 2.95 3.05 3.12 3.18 3.22 3.26 3.29 3.32 3.47 3.47 3.53
00 2.77 2.92 3.02 3.09 3.15 3.19 3.23 3.26 3.29 3.47 3.61 3.67

f= degrees of freedom.

Experimental Design
Cria43
11
3
 Table 5. Probability table P and Q for n= 5

Table II Binomial probabilities

n=5
050 .100 .200 300 400 .500
7k
0 774 .590 .328 .168 .078 031 5
1 .204 .328 .410 360 259 .156 4
2 .021 .073 .205 309 .346 .313 3
3 001 .008 .051 .132 230 .313 2
4 .000 000 006 .028 .077 156 1
5 000 .000 000 002 .010 .031 0

Nk
950 900 .800 .700 .600 ,500 TB

n=10
BT
kS .050 .100 .200 .300 .400 .500

0 .599 .349 .107 .028 .006 .001 10

1 .315 .387 .268 .121 040 010 9
2 075 .194 .302 .233 .121 .044 8
3 .010 .057 .201 .267 .215 .117 7
4 001 .011 088 .200 .251 .205 6
5 000 .001 .026 .103 .201 .246 5
6 .000 .000 006 .037 .111 .205 4
7 .000 .000 001 .009 042 .117 3
8 .000 .000 .000 .001 .011 .044 2
9 .000 .000 .000 .000 .002 010 1
10 .000 .000 000 .000 .000 .001 0
N
.950 .900 800 .700 .600 .500 TT \

Experimental Design
Cria43
11
4
Table 7. Student's t values

n=15

.050 .100 .200 .300 .400 .500

Tk
0 .463 .206 .035 .005 .000 .000 15
1 .366 343 132 .031 005 .000 14
2 .135 .267 .231 .092 .022 .003 13
3 .031 129 250 .170 063 .014 12
4 .005 .043 .188 219 .127 .042 11
5 .001 .010 103 .206 186 .092 10
6 000 .002 .043 .147 .207 .153 9
7 .000 000 .014 081 .177 196 8
8 .000 .000 .003 .035 118 .196 7
9 000 .000 .001 .012 .061 153 6
10 .000 000 000 003 .024 .092 5
11 .000 .000 000 .001 .007 .042 4
12 .000 000 .000 000 .002 .014 3
13 .000 000 000 000 .000 .003 2
14 .000 .000 000 .000 000 000 1
15 .000 .000 .000 .000 .000 .000
0
Nk
950 .900 .800 .700 .600 .500 T,

n = 20
NT 050 .100 .200 .300 400 .500
R
0 .358 122 .012 .001 000 000 20
1 .377 .270 .058 .007 .000 .000 19
2 .189 .285 137 .028 .003 000 18
3 .060 .190 .205 072 .012 .001 17
4 013 .090 .218 130 .035 005 16
5 .002 032 .175 .179 .075 .015 15
6 .000 .009 .109 192 124 .037 14
7 000 .002 .055 .164 .166 .074 13
8 .000 000 .022 114 180 .120 12
9 000 .000 .007 065 .160 .160 11
10 000 .000 002 .031 .117 .176 10
11 .000 000 .000 .012 .071 .160 9
12 000 000 .000 - .004 035 120 8
13 .000 000 .000 001 .015 074 7
14 .000 000 000 000 .005 .037 6
15 000 .000 .000 000 001 .015 5
16 000 000 000 000 .000 .005 4
17 000 .000 000 000 .000 .001 3
18 .000 000 .000 .000 000 .000 2
19 000 .000 .000 000 .000 000 1
20 000 .000 000 .000 000 .000 0

NE t
.950 .900 .800 .700 600 .500

Experimental Design
Cria43
11
5
Table 6. Probability table P and Q for n= 15

Student's t values and associated

probability P as a function of degrees
of freedom df.
F1
Yea 0.05
(of a cell)
0.4 0.25 0.1 0.025 0.01 0.005 0.0025 0.001 0.0005
h
2 0.289 0.816 1.836 2.920 4 303 6.955 9.925 14 099 22.326 31.596
3 0.277 0.765 1.638 2.353 3 192 4.54- 5.841 7 453 10.215 12.924
4 0 271 0.74- 1.533 2/32 2 776 3.747 4.604 5.598 7.173 9.61
5 0 267 0.727 1.4 2.01 2 571 3.365 4.0 4.773 5.893 0 6.86
6 0.265 0.718 1.440 1.94 2 447 3.143 3.707 4317 5.208 5.95
7 0 263 0.71- 1.4 1.89 2.365 2.998 3.499 4 029 4.785 5.40
8 0 262 0.706 1.397 1.86 2.306 2.896 3.355 3 833 4.50- 5.04
9 0 261 0.703 1.383 1.33 2.262 2.821 3.250 3.690 4.297 4.78
10 0' 0.700 1.372 1.81 2 228 2.764 3.169 3 591 4.144 4.58
260 2 1.79 7
11 0 260 0.697 1.363 2.201 2.7'8 3.106 3.497 4.025 4.43
12 0 253 0.695 1.356 6 1.78 2 179 2.831 3.055 3 428 3.930 7 4.31
13 0 253 0.694 1.350 1.77 2 160 2.650 3 012 3.372 3.852 4.22
14 0 258 0.692 1.345 1 2 145 2.824 2.977 3 328 3.787 4.-40
15 0 258 0.691 1.341 1.75 2 131 2.802 2.947 3 298 3.733 4.07
16 0 258 0.690 1.337 1.74 2 120 2.583 2.921 3 252 3.886 4.01
17 0 257 0.689 1.333 1.74 2 110 2.567 2.899 3 222 3.846 3.96
13 0 257 0.688 1.330 1.73 2.101 2.552 2 879 3 197 3.8'0 3.92
13 0 257 0.688 1.328 1.72 2.093 2.539 2.361 3 174 3.579 3.88
20 0 257 0.687 1.325 1.72 2.096 2.528 2.345 3 153 3.552 3.35
5 1.72 0
21 0 257 0.686 1.323 2.090 2.5-8 2.331 3 135 3.527 3.81
22 0.256 0.686 1.321 1 1.71 2 074 2.508 2.819 3 119 3.505 9 3.79
20 0.256 0.685 1.3'9 1.71 2.069 2.500 2.307 3 104 3.485 3.76
24 0.256 0.685 1.3'8 1.71 2.064 2.492 2.797 3 091 3.467 3.74
25 0.256 0.684 1.3 1.70 2.060 2.485 2.787 3 078 3.450 3
26 0.256 0.684 1.3 1.70 2.056 2 479 2.7 3.067 3.435 3.70
27 0.256 0.684 1.3-4 1.70 2 052 2.473 2.771 3 057 3.42' 3.69
28 0.256 0.683 1 313 1.70 2 048 2.467 2 763 3 047 3.403 3.67
23 0.256 0.683 1.3'1 1.69 2 045 2.462 2.756 3 038 3.396 3.65
30 0.256 0.683 1.3 1.69 2 042 2.457 2 750 3 030' 3.385 3.64
7 6
40 0.255 0.681 10
1.303 1.68 2 021 2.423 2.704 2 971 3.307 3.55
60 0 254 0.679 4
1.206 1.67 2 000 2.390 2.660 2915 3.232 1 3.46
120 0 254 0.677 1.239 1.65 - 990 2.358 2.617 2.860 3.160 3.37
Infini 0 253 0.674 1.232 1.64 ■ 960 2 326 2.576 2 807 3.090 3.29
te 5 1

Experimental Design
Cria43
11
6